LIBRA Jr.) to -1- I

Or LIBRA Jr.) to -1- I

APOSTATIC SELECTION: THE RESPONSES OF

WILD PASSERINES TO ARTIFICIAL POLYMORPHIC PREY

John Ashdown Allen

Thesis submitted for the Degree of Doctor of

Philosophy University of Edinburgh, December 1972 -2-

ABSTRACT

The iaintonanco of non-mimetic colour polymorphism has yet to be satisfactorily explained. Apotatic selection is one possible answer: sight-dependent predators may tend to form 'searching images' for common morphs and concentrate on those varieties in preference to rarer ones • (Chapter t).

Some evidence for apostatic selection exists, but it is mostly indirect and incomplete'. (Chapter II).

The hypothesis was tested by experiments with wild passerine birds in their normal surroundings. The prey wore lard-and-flour

'baits' which were usually green or brown and cylindrical (0.7cm long x 0.7cm diameter). Data were obtained by direct or indirect observation of 'populations' containing the colours in known proportions. Eaton baits were replaced frequently to ensure that the ratios wore kept constant. Five cots of experiments wore carried out. (chapter II!).

In one set (Chapter IV) baits were at maximum density. Mon unequal numbers of greens and browns were presented there was a significant overall tendency for the ftrer forms to be removed.

Individual blackbirds Tur'dua morula tended to concentrate on one colour alone within visits. Presentation of greens and browns in equal numbers gave no evidence that the colours differed in taste.

Further experiments with small green and brown baits and small rod and yellow baits confirmed that rare forms are preferred at maximum density. -3-

In the remaining sets of oxperimonte baits tiore at a density of tio per square metre. Populations of randomly distributed groans and browns were presented on grass or soil backgroncIe.

A second sot of axperiments (Chapter V) revealed that birds can bocoize conditioned to searching for either green or broin baits. Mid birds were fed with one colour alone and toro then given populations with both colours in equal proportions. In all eight operiments the familiar colour was strongly preferred. In three cases the preferences woro reversed by a further training period on the'unfamiliar' colour.

A third cot (Chapter VI) shoed that psscorinea on average tend to take more of the common variety than aEpected on the bcsi of constant selection • In each experiment populations tore presented either with greens nine times as comian as brome folloed by populations with brome nine times as common as greens or in the rerso order. The initial study revealed heterogeneity within the data and repeat experiments wora therofere carried Cut. The data frW the fourteen experiments wam consistent • In thirteen cases the common colour was preferred ovoll, despite different environmental conditions.

A fourth sot investigated the behaviour of individual blackbirds and songtbr'uahos Turdua philociclos. One experiment

(Chapter VII) showed that the three birds involved preferred brotmo • In two cases these preferences were very strong and the birds otartod to oat greens in quantity only tihen brww mere very rare • This behaviour woo preceded by an increase in the tendency to 'imprint-pew. The three birds appeared to differ markedly -'4 -

in their rasponsas. Too further eporirents (Chapter VIII) with a majority of 'nc& birds confirmed the above findings and also provided evidence of searching images. At the and

of the third experiment populations of green, browns and khstde were offered. The khakis were initially overlooked.

In a fifth set (Chapter IX) greens, browns and various shades of khaki were used. Two series of experiments tooted the

specificity of conditioning. Birds were familiarized with one

colour and were than offered populations with the 'familiar'

11 bait and a visually similar colour in equal proportions. They were then trained on the 'unfamiliar' colour and given a choice as before. The results showed that birds can become conditioned

to a particular colour; this conditioning can be very specific, but reversible • In two further experiments an attempt was made to see whether predators Can promote polyorphietm. Blackbirds and songthrusbes wore offered populations containing nine morphs ranging from green to brown and in proportions that fitted the normal distribution. These praportionQ were altered regularly to accord with the freuoncioa that z'oiaatnod after a fixed percentage of baits had been eaton • One eporiment showed that the browner baits were preferred. The second, experiment was preceded by a period of familiaritation with the cotxnzonoet khaki morph and showed an increase in the variance of the population.

The distribution became bimodal • This disruptive effect was probably a result of the training, natural brown preferences, and the heterogeneity of behaviour of the birds.. Searching images

can therefore be very specific and under certain conditions predators may be capable of causing polymorphism in prey species. -5--

The oxporixita1 dasign is conidosd to resemblo a natural situation. It is ocnàludod that the results support the hyothoio of apoetatic Ooli3ction • The findingo have Implications fbr the conditions under which spostatic selection is likely to act and point the way to futura rosGarch i It is ougostod that a no of different typea of opoetcxtie polymorphism odot. (Chapter X). -6 -

PREFACE

My aim in this work has been to test a hypothesis. It has been suggested that variability within a population of a palatable prey species can be maintained by 'epostatic selection' by predators, the meaning of which will become clear later. This posibility has been investigated by presenting populations of artificial prey to wild passer'ino birds in their normal surroundings. The exorcise is therefore oituatod in the area of zoology where animal behaviour and population genetics overlap. We are concerned with the behaviour of predators and more important, with its effect an variability in the prey population.

The thesis is organised into ten Chziptorc. The first is introductory and provides the background to the problem. Existing evidence for apostatic selection is reviewed in the second Chapter.

A description of the general experimental approach is given in the third and the experiments themselves are discussed in the following seven Chapters. Bulky data are tabulated in appendices at the ends of the relevant Chapters. The final Chapter focuses on the main conclusions and points out possible lines of future research.

Acknowledgements

First and foremost I thank my supervisor Professor Bryan Clarke for his constant encouragement, criticism and discussion throughout all stages of the work.

I am also grateful to many other friends and colleagues who have discussed aspects of the work with me. In particular I must mention those members of the 'population genetics group' who were - 7 - at-the Department of Zoology, University of Edinburgh, between

1967 and 1971. I am indebted to Professor J.M. Mitchison and

Professor P.M.B. Walker for the facilities that I have enjoyed in their Department. The research was supported by a Research

Studentship from the Science Research Coulicil, to whom I extend my appreciation. Various other people gave help and I thank them all. They include those who allowed me to work on their land: Professor

Bryan and Mrs. Ann Clarke; the late Mr. J. Ewing; my parents, Professor and Mrs. P. Allen; and the farmers of Midlothian who allowed ma access to their fields. I am especially grateful to Miss Nan Brownlie, Mr. Kim Howell, Mrs. A. Almeida and Mrs. A. Patel who typed the first draft, and in particular Mrs. D. M.Powell who typed the thesis in its present form. The final typing, assembling and binding was done in Reading while I was in Dar es Salaam and I am very much indebted to Mrs. Powell for handling the arrangements at that end. My wife, Eleanor, gave much appreciated assistance in checking the final typescript. Mr.

Bob Brown photographed most of the Figures and Mr. David Briscoe

photographed the Colour Plates. Finally, I thank a motley crowd of people without whose

spiritual support this thesis would never have been finished. These

include Eleanor, Boris and Chris, John Clarke, Bones, Cohn Craig,

David and Jane, Captain Beefheart, and latterly Michael, John and Jan in Dar. All these people kept my sanity at times when prospects

seemed bleak.

-8-

CONTENTS Page

ABSTRACT 0 ,0 . 0 . • oo.o.... 000 000 2

PREFACE ..... 00006 ...... 0 ...... 6 Acknowledgements ...... 6 TABLESTEXT -FIGURES, PLATES ...... , ...... 13

TABLES. 0 0 0 0 00 0 0 ...... 00 ...... 0 0 0 00 0 4 13

TEXT-FIGURES ...... 0 0 0 • , • • • . . . • • • ...... 15

3 • PLATES ...... • • • • • • 0 0 0 0 0 0 0 0 4 ...... 16

CHAPTER I GENERAL INTRODUCTION ...... 17

1. COLOUR POLYMORPHISM AND ITS MAINTENANCE ...... o.... 18

2. SELECTION BY PREDATORS , ...... • 0 • • ...... 21

General 0 • • • • , • 0 ...... • , • 0 0 0 0 0 0 0 0 0 0 0 0 • 0 0 00 0 • 0 0 21

Apostatic selection • • • • • • • •• • 00 0 0 0 0 24 3. APOSTATIC SELECTION AND SEARCHING IMAGES 26

Searching image 0 0 00 • • 0 • • • ...... 0 • • • • ,0 26

Searching images and colour polymorphism 0 00 00 28

CHAPTER II THE EVIDENCE FOR APOSTATIC SELECTION ,.o....,, 29

1 . IN'rIDtJc'rIoN 0040000 • ...... , • 0 • • 30 2 . MIXED COLONIES OF CEPAEA ...... 31

3. PREDATOR-PREY STUDIES ...... 0 ...... • • 32 (a) Introduciñg the evidence ...... 32 (b) Do predators become conditioned to searching for specific palatable prey? ...... 33

( t) General evidence ...... 000 33 (ii) Natural polymorphic prey •...... 38 (c) Do predators become conditioned to searching for the commonest types in a mixture of palatable prey ...... 141 General evidence ...., ...... , ...... ,, 41 Natural polymorphic prey ...... , 46

4 . CONCLUSIONS • • • • • • 0 0 0 0 0 0 • • • ...... '47

(a) Conditioning to a specific prey •...... 000 47 Specificity of conditioning ., 47

Apostatic selection 0 ..... 00 000 0000 ...... 48

CHAPTER III GENERAL MATERIALS AND METHODS ...... 00 ...... '49

INTRODUCTION ...... 000 • •. • ...... •. .. . 50 CHOICE OF EXPERIMENTAL SYSTEM ...... ,....., 50 BIRDS •.,...,...... 00000*0000 ...... , ...... 51 4 , PREY ...... 0 ...... 53 Morphology ...... 0 0 .....0 0 0 00 ...... 0 0 • • 53 Manufacture ...... 35 ENVIRONMENTS ...... 0 0 ...... 55 TECHNIQUES ...... 0 0 0 0 ...... 0 0 0 0 • 0 0 0 0 ...... 56 (a) Presentation ofprey ...... 56 (1) Experimental plot ...... 56 (ii) Randomization of prey distribution 58

-9-

Page

(b) Recording of data ...... 60 7. PLAN OF EXPERIMENTS ...... , . . 61

CHAFFER IV MAXIMUM-DENSITY EXPERIMENTS ...... •... 64 1.. INTRODUCTION a . . , a • . . . a . , . , . , . ...... a o a a a a a a • 65 2. MATERIALS AND METHODS ...•...... 67 Series 1 ...... 67 Series 2 , ...... 67 3 . RESULTS . , • * • , , , • ...... 70 Series 1...... ,., 70 ( i) Synopsis ...... ,..., 70 (ii) Behaviour of birds .., ...... , 72 Series 2 . . . . . * ...... 74 4 . DISCUSSION ...... . • ...... 78 Colour-preferences and searching image ., 78 Së1ectionof rare prey ...... 80 5 . SUMMARY ...... 82

CHAPTER V CONDITIONING TO GREES OR BROWNS ,,...... , 84 1. INTRODUCTION . • ...... • . . • 85 2.PRELThINARY EXPERIMENT ...... 1 ...... 0 85 Procedures...... 85 Results and discussion .,...... , 87 S . REPEAT EXPERIMENTS . . . . . -...... 87 Plan ...... , . a 87 Materials and methods ...... 88 (1) Locations . . . 4 • , • , • • • • • • ...... 88 (ii) Procedures •.*., ...... 88 (ill) Predators ...... ,.a, 92 Results ...... • * •, • *• ...... a...... 92 4. DISCUSSION ...... * ...... 96 Comparison between preliminary and repeat experiments ...... 96 Effect of natural preferences 96 Conditioning .. . . , ...... , . . . . , ...... 97 5 . SUMMARY ...... ,...... •, ...... 100

CHAPTER VI 9:1 EXPERIMENTS: PREDATION BY GROUPS OF BIRDS...... * .. * . . . 101 INTRODUCTION ...... , . * . . . * ...... * ...... 102 PRELIMINARY EXPERIMENTS • ...... 102 Experiment 1: presentation of 9G:1B populations followed by 1G:9B populations. 102 (1) Materials and methods...... ,. 102 (ii) Results and-discussion ...,...,..... 104 Experiment 2.1: presentation of 1G:9B populations followed by 90:1B populations. 108 Materials and methods ...... 108 Results and discussion ...... ,,., 108

- 10 -

Page (c) Experiment 4: presentation of 1G: 9B populations followed by 1G:1B and 9G:1B populations ...... 110 (1) Materials and methods ...... ,..., 110 (ii) Results and discussion ...... , 110

3 . REPEAT EXPERIMENTS • • • 0 0 0 0 0 0 0 • 0 00 • 0 0 0 00 0.0 111

(a) Plan of work • • , . , . , ...... • * 0 * • • • , • 112 (1 ) Series 3 . . . • • • , • , . . * . • • • • • • • • 112

Series 5 . . • • • 0 • • 0 0 0 •, ...... 1.12 Series 6 ...... 113 (b) Experiments and results •...... 113

4. SUMMARY OF DATA ...... 0 0 0 ...... 113 S. DISCUSSION , ...... , . 116 Apostatic selection ...... , 116 Colour preferences ...... 118

Conditioning •,.0000...... , 121

6. SUMMARY ...,,...,....,,...... 00000000000 122

APPENDIX A: TABLES 22-32 0000 ...... ,...,...., 123 (a) Key to Tables ,, ...... • ...... 123 (b) Tables and expsrintsnts 123 (1) Presentation of 9G:1B populations followed by lG:9B populations 125 (ii) Presentation of 1G:9B populations followed by 9G:LB populations .,. 137

CHAPTER VII THE RESPONSES OF INDIVIDUAL TURDIDAE:

PART 1 0 a , ...... 0 • 0 o . . . . 146

INTRODUCTION 0 • 0 • 0 0000•e0 • • • • , • • • • • 0 , • •- . . 148 EXPERIMENTAL AREA AND BASIC MATERIALS AND • METHODS ..... ,,,....,...... ,,,.,.,...,..., 148 • (a) Experimental -procedures ...... ,.. 148

(b) Birds ...... 0 0 • •, 0 0 ...... , • • • • • • • 0 0 0 , 149 (c) Environmental variables .....,...... ,,... 150 (i) Temperature ...... 150 • (ii) Available natural food ..,....,..... 150

(iii) Rain 0 • • • • * . . • • • • • • • • • • . . . . . 150

(iv) Grass ...... • 0.0000000.,.., ...... 151 3. EXPERIMENT 2.1, 1G:9B POPULATIONS FOLLOWED BY • 90:1B AND OTHER 'GREEN-DOMINATED' POPULATIONS. 151

(a) Chronology ...... • • • 00 • •• • • • • • • 151 • (b) Materials and methods ....,..,..,..,.,,... 151

( c) Results ...... • • * • • • • , • 0 0 • • • • • • • • 0 0 153 (i) Summary of data ...... ,. 153 (ii)Apostatic selection ..,.,...... 154 (iii) Relaxation of brown preferences .... 156 (d) Discussion ...... •....,,.,.,..,...,..• 157

(e) Conclusion . . . S • • • • • , • • , ...... • • , . • . 159 APPENDIX B: TABLE 36 ....,., ...... 160 CHAPTER VIII THE RESPONSES OF INDIVIDUAL TURDIDAE:

PART 2 ...... * ...... • 0 0 • • • • 0 0 0 162

INTRODUCTION . . , 0 ...... • • • • • • , , • 0 0 • 0* .. 163 EXPERIMENT 2.2: 9G:lB POPULATIONS. EXPERIMENT 2.3 a b: 9G:1B POPULATIONS FOLLOWED BY 1G:9B POPULATIONS ...... 163

-11-

Page

Chronology . ...... • 4 • • • * • • • . . . . , 163 Materials and msthods...,...... , ...... 163 Results . , . . . . . , ...... ...... 165 (i) Summary of data •...... 165 (ii) Predation ..,..... *,••009005050090.00S • 166 (III) Imprint-pecking ....,...... ,. 170 (iv) Short-term preferences ...... 172 3. EXPERIMENT 2.3 (CONT.). PARTS c, dand a: TRIMORPHIC POPULATIONS 174 Chronology ...... •099000 175 Materials and methods ...... , 175 Results ,00.Oo00000.,o.O,,50.000.,,.D .... ,.... 175 (1) Summary of data .. . •... . • ...... * • 9 175 (ii) Predation ••,.•...... ,...,.,...... ,.,.. 177 4. DISCUSSION ...... * ...... * . . . . . 177 (a) Possible causes of colour preferences • .. .•. 178 (1) Relative conspicuousness ...... ,.,....., 179 (ii) Taste ..**.• ,...*...... 180 (iii) Innate preferences ...... 180 (iv) Permanent acquired preferences ...... 180 (v) Temporary acquired preferences ...... 181 (b) Searching images . ...... •. •...... 181 (c) Relevance of imprint-pecking ...... 183 (d) Diversity of behaviour and its implications,.. 185 (i) Diversity of behaviour ...... 185 (ii)Apostatic selection ,...... 187 (iii) Predators and area effects ,...... 190 (a) Introduction of third morph ...... ,...... 193 S. SUMMARY OF CONCLUSIONS FROM EXPERIMENTS IN CHAPTERSVII AND VIII ...... , ...... • . . 193 APPENDIX C: TABLES 42 AND 43 ...... 193

CHAPTER IX FURTHER EXPERIMENTS WITH GREENS, BROWNS AND INTERMEDIATECOLOURS ...... 197 1 . INTRODUCTION . . . , ...... 198 2 . PREY *500•*4*öO*OO..5.*.*..O.*.*D*O9..D*90..*O5*.O 200 3. EXPERIMENTS 1 AND 2 : TRAINING ...... ,.. 202 (a) Introductory remarks ...... ,,.,...... 202 (b) Plan , • ...... , . . . , ...... , . . , . . . . , . , . . . . . 202 (c) Materials and methods ...... 204 (1) Locations ...... ,.,.,...... ,,...... 204 Procedures . . • . . ...... . . . . . . 204 Predators ...... ,..,....,...,. 204 (d) Results ...... 206 4. EXPERIMENTS 3 AND 4: 'NORMALLY-DISTRIBUTED' - POPULATIONS ...... 208 Introductory remarks...... 208 Basic design of experiments ...... 211 (i) Prey 211 (ii) Presentation of populations ...... 211 (Iii) Changing the compositions of the populations 9SS***•* •*bO0••O• O.....,... . 212 (a) Practicability of methods ...... 213 - 12 -

Page

Preliminary study: experiment 3 ...... 215 (1) Materials and methods ,...... ,...... 215 (ii) Results and discussion ...... 215 Experiment 4 ...... . . . . 221 U) Materials and methods .. .. ,...... 221 (ii) Results and discussion ...... 223 5 . DISCUSSION ...... 228 Stimulus generalisation and the specificity of conditioning •.... .. .. •...... 228 Disruptive selection ...... ,.....,...,...... 249 (1) Maintenance of distinctness of morphs . 251 (ii) Maintenance of variability ...... 253 6. SUMMARY . . • ...... 6 • • . 254 APEENDIX D: TABLE 54 . , ...... 255

CHAPTER X GENERAL DISCUSSION AND CONCLUSION ...... , 258 1. WILD PASSERINES AS PREDATORS OF POLYMORPHIC SPECIES...... * ...... 259 2 • BAITS AS NATURAL PREY ...... • • • • • • . . . • . . . 261 30 EVIDENCE FOR APOSTATIC SELECTION ...... , 262 Conditioning to a specific prey •. • • • 262 Specificity of conditioning • ...... 262 (c)Apostatic selection ...... 4. CONDITIONS FOR APOSTATIC POLYMORPHISM .....,.., 266

Density . • ...... I • • 6 0 • ...... 268 Conspicuousness ...... • . . . . . 271

Palatability ...... a ...... 274 S. TYPES OF APOSTATIC POLYMORPHISM ..,...... 275

6 . CONCLUDING REMARKS ...... 0 0 • • ...... 280

REFERENCES ...... • • • • . • • • ...... • ...... 281

APPENDIX E: PUBLISHED PAPERS . .. 0..O666•6 ... 294 - 13 -

TABLES, TEXT-FIGURES, PLATES

1. TABLES Page qpter III: 1. List of birds in experiments ...... 52 to 2. Colours of green and brown baits ...... 52 Chapter IV: 3. Colours of red and yellow baits ...... 68 4. Grand totals of baits taken in

experiment 1 ...... 60 ...... 69 5, Numbers of baits taken by blackbirds during observed visits in experiment la 71 6. Numbers of baits taken by blackbirds and house sparrows during observed visits in experiments lb and ic ...... ,...... 73 7. Grand totals of green and brown small baits taken during experiment 2a ...... 75 8. Grand totals of yellow and red small baits taken during experiment 2b ...... Chapter V: 9. Preliminary experiment. Daily totals of it baits taken after familiarization ...... 86 10. Sites of experiments in series 1, 2, 3 .. 89 11. Dates of training and presentation of 1:1 populations in series 1, 2 9 3 ...... 90 12. Numbers of baits taken from 1:1 populations in series 1, 2, 3 by blackbirds and house sparrows •...... 91 It 13. Experiments 3,1, 3.2 : daily predation byblackbirds •0* • ... •...,.....•... 93 14. Numbers of baits taken by recognisable female blackbird in experiment 3.1 ..... 95 Chapter VI: 15. Experiment 1. Daily totals of baits taken . . , • , • ...... , , . . . , ...... 103 16. Experiment 1. Numbers of baits taken by blackbirds during individual visits..... 105 17. Experiment 2,1. Daily totals of baits taken ...... 107 tt 18. Experiment 4. Daily totals of baits taken 109 19. Summary of data from 9:1 experiments

(9G:1B + IG:9B) •.000.000•o ...... 114 20. Summary of data from 9:1 experiments (lG:9B -' 9G:1B) 115 21. Percentage browns taken on first days of predation on 9G:1B populations...*.*...* 120 Appendix A:22-32. Results of repeat 9:1 experiments ...... 123 to 22 Experiment 3.1 . . •o•oe.•.•0 •.oD.oeo .. . 124 it Experiment 3.4 ...... 126 of Experiment 5.2 ...... ,. 128 It Experiment 5.3 . . ...... 130 IT Experiment 5.4 ...... •...... 132

Experiment 6.2 . .. 0•O0O ...... ,. ..,.,... 134 Experiment 3.2 ...... 136

29 0 Experiment 3.3 ...... * ...... 138 Experiment 5.1 ...... • ...... * • • • .. . 140 Experiment 5.5 ...... 142 IT Experiment 6.1 .....,...... •...... 144 * 14 —

Page

Chqpter VII: 33. Experiment 2.1. Grand totals of baits taken and imprint-pecked by blackbirds A , B, and songthrushT •...... 152 it 34. Predation by B in experiment 2.1 ...... 154 ft 35a, 35b, 35c. Behaviour of A, B, T, when starting predation on greens ..,...... 155 Appendix 8: 36. Experiment 2.1. Full daily results for A, B , T ...... * ...... , ...... 161 Chapter VIII:37. Experiments 2,2a and 2.3a,b. Grand totals of baits taken and imprint-pecked by blackbirds C, D, E, C, and songthrush T ...... 164 it 38. Behaviour of C in experiment 2.2a ...... 167 it 39. Proportionlaaten and imprint-pecked by C, D, E, C, T in 2.2 and 2.3a ...... 169 it 40. Sequences of like and unlike baits taken by C, D, E, C, T in 2.2 and 23b ...... 173 of 41. Experiments 2.3c,d,e. Trimorphic populations : grand totals of baits taken and imprint-pecked by blackbirds C, D, E, G, I, robins, dunnocks, house sparrows .... 176 Appendix C: 42. Experiment 2.2. Full daily results for C, D , E, C, T ...... •.o.... 0*e*t ...... ,,. 195 43. Experiment 2.3. Full daily results for C, D, E, C, I, robins, dunnocks, house Spx't'OWs • .....• a a a . a • • a a a a a a a o . a a a a a a 196 çpter IX: 44. Colours of nine types of green, brown and khaki baits ...... , , . * ...... 20]. Sites of specificity training experiments . 203 Dates of training and presentation of 1:1 populations ...... • . ...... . . . . 203 II Grand totals of baits taken from 1:1 populations after training ...... 205 48a,Specificity training experiments, series 1. Daily results for blackbirds and sinai]. birds ...... . . , , . . . a • , • • • . 209 48b.Specificity training experiments, series 2. Daily results for blackbirds and small birds . . . . . . . , ...... , ...... 210 49, Initial composition of 'normally- distrThuted' populations ...... ,...... 2]1 50'. Experiment 3. Composition of whole populations in each of 3 generations ...... 214 Predation by male and female blackbirds

in experiment 3 6*a • ...... ,...... 218 Experiment 4. Composition of whole populations in each of 20 generations ..... 222 Mean colour of baits taken by blackbirds E, J, K, L, M and songthrushes T's over every 3 generations in experiment 4 s...... 227

- 15 -

Page

Appendix V: 54. Number of baits taken over every 3 generations by E, J, 1<, L, M, and In experiment 4 ...... 256

2. TEXT-FIGURES

Chapter I: 1. Frequency of a morph in population plotted against frequency predated by • an'apostatic'predator 23 Chapter II: 2. Relationship between density of Acantholyda nemoralis and its proportion in tits' diet (from Clarke 1967a) 40 Predation by rudd on 2 and 3 varieties • of corixid bugs (from Clarke 1962a) 43 Predation by minnows on dimorphic population of artificial prey (from Smith

1967) ...... 0400.0 .....•• ...•...... 45 'Roneo'-copy of grid to show distribution of baits ...... 57 Chapter III: 6. Interrelations of Chapters IV to IX 62 Chapter IV: 7. Experiment 2a. Percentages of small greens offered plotted against percentages taken 76 it B. Experiment 2b. Percentages of small reds offered plotted against percentages taken 76 01 9. Percentage4yeliow seeds offered plotted against percentages taken (from Pough 1964) . * ...... , , ...... , . . . . . 79 Chapter VII: 10. Sites of plots in experiments 2.1, 2.2,

2.3 •.,e..o.o .... 0000.,.g,.0..e..o ... 00000 147 Chapter IX: 11. Experiments 1.1, 1.2, 2.1, 2.2. Percentages of 4's taken over every 5 visits from 1:1 populations after specificity training 207 It 12. Experiment 3. Morph frequencies in population before and after 3 generations

of selection 46000.e.,,..,....000.s..e,000 216 13. Site of plot in experiment 4 .., 220 01 14. Experiment 4. Morph frequencies in population after every fifth generation of selection . . . . . * ...... 224 re 15-21., show the numbers of baits taken by each bird in experiment 4 and the average composition of the population over: it 15. generations 0-2 ...... 231 it 16. generations 3-5 ••• .00.00000.0 ...... 233 It 17. generations 6-8 ... , ...... 235 it 18. generations 9-11 ...... 237 ti 19. generations 12-14 ...... ,...... 239

generations 15-17 ...... 0000 ...... 241 generations 18-19 .....,.....,..., 243 - 16 -

3. PLATES

Page

Chapter III: Standard brown and standard green

baits •Gt ... 0..O*0000tOG.000..OG.O 54 IV: Maximum density experiments. 1:1 populations of standard greens and standard browns ...... * . ...... 66 Maximum-density experiments. 1:1 populations of small greens small browns, and small reds, small

yellows ...... ...... 66 çpter IX: Appearances of the nine types of green-brown baits ...... 199 - 17 -

CHAPTER I GENERAL INTRODUCTION - 18 -

CHAPTER 1 GENERAL INTRODUCTION

1. COLOUR POLYMORPHISM AND ITS MAINTENANCE

Individuals of some species can be classified into a number of distinct forms living in the same population at the same time. Then the rarest 'morph' is present at a frequency too high to be accounted for by mutation, this variation is defined as 'polymorphism' (Ford

190). It usually has a genetic basis and may be detected at different positions a1onga biochemical pathway; from close to the action of the gene, as in enzyme polymorphism, to actually visible

alterations of morphological characters. At this end of the system, the colour or colour-patterning of the organism may be involved. Some colour polymorphisms (as perhaps in Biston betularia L., Kettlewell 1958) may only be transient, with one morph gradually

replacing another. Others appear to be relatively more stable. This latter, balanced, type poses the more interesting evolutionary problems, in particular with regard to how it is maintained. The possible mechanisms capable of maintaining genetic polymorphism

have been examined theoretically by a number of authors (e.g. Dempster 1955, Sheppard 1958, Williamson 1958, Maynard Smith 1962, Ford 1965). In practice, however, the relevant selective forces have rarely been

identified. One possible exception concerns polymorphic mimicry in

certain tropical Lepidoptera (e.g. see Clarke and Sheppard 1960 for

Papilio dardanus Brown, Ford 1953 for Hypolimnas misip L.). In these, some or all of the morphs mimic other more distasteful species. A mimetic morph will be favoured so long as its frequency relative

to the model remains low. Once the frequency of the morph increases

its advantage starts to diminish. A definite disadvantage will

occur when it is so common that its conspicuous coloration - 19 - becomes associated with palatability rather than the opposite. Other morphs may therefore be at a relative advantage and a balanced polymorphism can result. However, there are many examples of colour-polymorphism where mimicry is not clearly involved. They are to be found throughout the

Animal Kingdom. Because of lack of space it is impossible to give a complete or even representative, list and no comprehensive review exists. It is hoped that the following examples give some idea of the range of species involved:

Metridium senile (Coelenterata, Anthozoa), Fox and Pantin (1961).

Pomatoceros trigueter (Annelida, Polycheeta), F$yn and Cjden (1954). c!paea spp. (Mollusca, Gastropoda), various authors. phaeroma rugicauda (Crustacea, Isopoda), West (1964). Arinadillium nasuturn (Crustacea, Isopoda), Adamkewicz (1969).. Philaenus spp. (Arthropoda, Hemiptera), HalJka and Lallukka (1969). Homorocoryphus nitudulus (Arthropoda, Orthoptera), Owen (1965c). phiocemina nigra (Echinodermata, Ophiuroidea), Fontaine (1962). phphorus maculatus (Chordata, Pisces), Gordon and Gordon (1957). Eleutherodactylus spp. (Chordata, Amphibia), Coin (1950). Malaconotus spp. (Chordate, Ayes), Hall etal. (1966).

Colour-polymorphic species inhabit a diverse range of environments, both marine and terrestrial, and their distribution is world-wide. The degree of polymorphism varies from species with two morphs (e.g. the larvae of the moth, Geometra papilionaria L.) to species possessing almost infinite variation (e.g. the butterfly clams,

Donax spp.). Within species, there are often very distinct visible differences between the morphs. Indeed such variation has in the past created taxonomic confusion. Until recently, the different - 20 -

phases of the bush-shrikes (2a1aconotue) wore considered as separate species, simply because they looked so different (Hall at at. 1966).

Similarly, the 170 African species of the land-snail Limicolaria have now been reduced to 17 (Crowley and Pain 1970).

How, then, are those non-mimetic polymorphisms maintained? Faced with a lack of evidence for one mechanism or another, many workers have assumed that hetorozygous advantage is involved.

Documented examples of ouch views pertain to Copaaa app. (Cain and

Sheppard 1954a, Ford 196 1aeonotus app. (Hail ot'Ql. 1966) and the platyfish Xiphqphorus maculatus Gnthor (Gordon and Gordon 1957). It is true that there is definite evidonca that heterosia can maintain a polymorphism (for instance in the classic case of sick1ecoU anaemia, Allison 1936), but its ubiquity as a mechanism has yet to be demonstrated. With regard to colour' polymorphism, this lack of positive evidence may, admittedly, often be duo to the difficulties of distinguishing between homozygotes and heterozygotes in the wild. This is especially true for organisms, like Cepaq, where there is still ho satisfactory method of distinguishing the dominant genotypes.

Heterosis is by no means the only mechanism capable of maintaining a polymorphism. Other possibilitas have bean discussed in detail by Williarson (1958) and other authors previously mentioned (P-18)- Nora specifically, Clarke (1962a) has examined these mechanisms with regard to colour-variation, particularly in COpSQQ. He concluded that the colour and banding polymorphisms of these species ore probably maintained by selective predation, in the absence of mimicry. This thesis focuasa On such a mechanism. -21-

2. SELECTION BY PREDATORS

(a) General Before considering this possibility in detail, it is instructive

to examine some more gener'al aspects of the relationships between predators and polymorphic prey • Tha can best discuss these features with reference to the well-known land snails Copasa nomoralia L. and C. hortensis !1u11. These spa cioc are sympatric over a largo part of the British Isles and inhabit both grasslands and woodlands.

They are highly polymorphic for the colour and banding patterns of their shells, and the genetics of these characters have bén worked out (Cain, Xing and Sheppard 1960; Cook 1967; Cain, Sheppard and Xing 1968), In Britain, the main predators of Cepeos are the songthrush

Tuz'dus ph ilonielos Hart., various rodents, and rabbits • The most relevant to our problem are the former, since they hunt almost exclusively by sight, whilst small mammals are more dependent on olfactory cues. In addition, the thrush definitely possesses colour vision (van Eck 1939), whilst rodents and rabbits discriminate only in terms of black and white (UaUs 1942). Hence, mammalian predators selectively predate those morphs of C2gaea that contrast in tone with the background (Cain 1953) • It follows that the colour of a morph taken by a rodent is not necessarily the same as that attacked by a thrush under the same conditions.

Selection of palatable animals contrasting with their background is now well-accepted (see Cott 190) despite the assertions of earlier workers like Selous (1909) and Pearl (1911). Experiments on the moth, Biston betulaz'ia, are classics in proving that conspicuous palatable prey are picked out in preference to those harmonizing with - 22 -

the background (Kettlewell 1955) • Balanced polymorphic species are also subject to strong visual selection. The colours of -

C • nemoralis morphs often correspond with the coloration of the background, due to selection by thrushes (Sheppard 1951). A similar effect occurs in C. hortensie, though the correlation with the background may be due to a different genetic response by the

snail; for example, the commonest morph found on leaf litter in

beech-woods is not unbanded brown as in C. nemoralis, but five-banded yellow, with all the bands fused (Clarke 1960). Colour-polymorphisms

in other species are also subject to changes in morph-frequency,

probably due to visual selection (e.g. in water-snakes, liatrix app.,

Cam.tn and Ehrlich 1958; in Limicotax'ia martensiana, Owen 1965a). If polymorphism in Copaeia is maintained by non-visual effects,

then years of selection by predators should have produced the evolution of a uniform shell-colour appropriate to the colour and texture of the environment. Yet the morphs are still distinct in appearance, and have been for at least thousands of years.. Sub-fossil

Cagaea shells still reveal some of the discontinuity of the variation they possessed when alive, since the bands remain discernable (Diver

1929). Similar' evidence for the durability of colour-pattern polymorphisms has been found in other gastropods, for example, LiniCola±'ia (Oven 1966b) and 'felacuinantus (Ewers 1966). If sufficient specimens of true fossil polymorphic gastropods exist, the banding variation might still be visible. Distinct patterns can be retained on specimens as ancient as Devonian trilobites (Esker 1968) and even older Caphalopods and Bvzchiopode (Foez'ote 1930).

It is therefore tecpting to suggest that some typo of predator'- controlled mechanism is maintaining çpaea and other non-mimetic piymorphiams. t10 must now consider what form it is likely to take. - 23 -

-o a) I-

> U a) 3

Frequency in population

:iura 1 The fzt3qusncy of a aorph in a population plotted against the froqusney predated. The broken line shows the expected relationship if predation is random. A hypothetical ote!nplo of apostatic selection is given by the continuous line; the morph is over-predated at high frequencies and under-predated when rare • Note that this simple example doos not involve additional frequency-independent selection; the sigmold curve is symmetrical about the discontinuous line. - 2'. -

Cain at al. (1960) otatod that oco iorphe of Sovaca s oom to zsonb10 coipononto of their habitat and are therefore kept distinct

by predatoro. 0tan (1966a) haü presented a similar cao for a

polymorphic East Africanspecies - of graochoppor9 HcmorocoEMh uo nitidu1u. It was realised by Cain at al., hoover, that boloctian

on this baois could not m aintain a po1ytorphim, sinco it vould be

freu.ency-indopendont. In addition it has been argued that oco

morpbs (o.. bright pinks of çpaoa and pallids of I4tnicolearia

martonsiana) are con spicuous against all backgrounds (Ciero 196a, 1969; Otion 1960).

(b) ~goatnticsoloctionn

It is noi opportune to proeent a raothd of oolo ction by predato that may be capable of maintaining po1yiorphiero in the absence of aiaiczr. Theoretically, the system to oicplo. It rolico on the

protaico that predators preforontiaily ooarch for, and oat, the most conon morphs of a polymorphic s pecies. Under such conditions the

rerer fores, oven if relatively conspicuous, ohould be protoctad from predation and honco be maintained in the population (sea Fig. 1). The m.ch&nisn, to zeney*depndont since the fitnoso of tho onotypo (and therefore the gone) is invorcoly related to Ito frequency. That this type of selection can maintain a polyorphioin bee been demonstrated many times (Wright 1948 0 Li 1962, Heidco and Jayekctr 1963, Clarko and O'Doneld 1964). In this dissertation, the tom

apoetatio oolooUon' (Clarke 1982a) will be ueod for frequency. dependent selection that maintains a colourpattorn pólyophica, since ouch a uochanioci promotoc 9clpOOtaoy 9 . That is, it favouro those rorpho, or l apostatool that are vioually different froi the majority (Clarke 1950 1 .

'Thpoctasjo r0001965) is oynonyouo with 'apiostatic cóloction . - 'Apostatic polyaorphio' (Clarke 1962a) is pelyiorphioo eaintainodby apoetatic ooloction. -25-

Historically, the hypothesis was originally suggested by the

eminent naturalist Poulton (1884) to explain dimorphism in the

caterpillars of the Large Emerald Moth, Coometra 12apilicnaria. L.

The idea was first applied in relation to CeDaGa by Cain and Sheppard

(195a), whilst Maidens (1955) also suggested that birds have to learn that particular,2Maea visual patterns are signs of edibility and hence maintain the polymorphism. More recently, Clarke (1962a,

1962b, 1969) 6 has taken up the cause that apostatic ealoction is

maintaining the shell polymorphism of Caj2aoa, Since then, the hypothesis has been invoked to explain many other cases of colour-

pattern pdymorphisme (e.g. Limicolazia, Owen 1963a, 1968b, 1965a; a1aconotus, Owen 1967; Philaenus, Owen and -Yeigez't 1962, Helkka and Lailukka 1969; Monte, Safriel 1969; poUua, Schi$tz 1971). Payne (1967) has advanced the hypothesis that polymorphism in

cuckoos and raptors minimises the chances of recognition by host and prey species. If this were true, apostatic selection would be

acting 'in offence' Rarer morphs would be at an advantage over commoner forms because they would tend to parasitiso or destroy disproportionately more victims.

Moment (1962) pointed out that predators may be responsible for the nearly infinite variation found in, for instance, some brittlestar (Ophiuroidea) and Butterfly Clams (Donax app'.). In other words the population as a whole benefits because there is no repeatable visual cue for a predator to learn. Moment called this type of selection 'reflexive' because "the frequency of any one typo is dotenninod by a feedback relationship with all the other typos". Howevert it is difficult to speculate how ouch a polymorphism could evolve (apart from invoking group selection, Uynno-Edwards 1962), other than by frequency-dependent selection This fact seems to have - 26 - been realised by Li (1962) when reinforcing Moment's hypothesis with mathematical evidence. Colour-pattern polymorphisms, then, may bestow an advantage to rare morphs (Clarke 1962a) or to the population as a whole (Moment

1962). In more general terms, such discontinuous variation may be

classified as a type of 'protean behaviour' (Chance and Russell 1959, Humphries and Driver 1967, 1970) which proposes the existence of patterns of erratic behaviour whose function is to confuse predators. To the extent that the polymorphism is unsystematic, predators will have to learn that each morph represents food. The behavioural process of how they are believed to do this is known as the formation of a 'specific searching image'.

3. APOSTATIC SELECTION AND SEARCHING IMAGES

(a) Searching image Croze (1967, 1970) suggests that hunting by 'searching image', a term coined by a naturalist (von Uexkull 1934), is more or less synonymous with the experimental psychologists' 'choice from sample'

(Kohts 1923) and the human psychologists' 'matching by set' (Vernon 1952). These specialists have apparently named the same 'universal perceptual concept' • The common processes involved consist of three

successive stages (Croze 1967, 1970): (i) 'seeing' the sample, (ii) examining the choice situation, (iii) reaction to the matching

stimulus. These three components can best be illustrated with reference to ourselves, for searching images are continually used in our daily

lives. We become conditioned to looking for visual clues associated

with some kind of reward. For example, a blue jug on the meal-table - 27 -

may come to be connected with the idea that it contains milk, if in

past experience, it has always been filled with this liquid. This

psychological conditioning is equivalent to (1) above; seeing the

sample. It is only when a choice situation (ii) is presented that the previous acquisition of a searching image becomes apparent. This may arise when, say, two jugs of milk are presented on the table; one the blue, as before, the other a jug of a totally unfamiliar design. Provided that the reward cannot be directly seen in either

jug s the 'predator' will almost invariably take the familiar blue one; he is reacting to the matching stimulus (iii). The choice so made must be induced by the searching image per se, even if it is contrary to unlearned, innate preferences.

Von tJexkull (1934) was one of the first to recognise the possible existence of searching images, when he found that toads (Bufo bufo L.), fed on earthworms, subsequently snapped at matchsticks in a choice situation, and at ants or pieces of moss, after a diet of spiders. The speed with which a searching image can be formed, and its specificity, is exemplified by the work of de Ruiter (1952) on the effectiveness of camouflage in the stick-caterpillar, Ennomos al can era L. Caged hand -reared jays (Garrulus glandarius L.) and chaffinches (Fringi].la coelebs L.) were used as predators. When sticks were added to the cages, the birds initially took interest in these new components.of their environment and pecked at them.

After the birds had become habituated to the twigs, dead caterpillars were also added. For a long time these went unnoticed, but once one was found (often by a bird accidentally treading.on a caterpillar), others were also discovered. Initially, in,their hunt for larvae, caterpillar-like twigs were also pecked, but the birds quickly became more efficient. This was even true for those experiments where the sticks were from the specific food-plant of Ennomos. Thus a searching - 28 - image can be extremely specific. As de Ruiter points out, however, birds in the wild would not have such a limited hunting area, and would have more time to habituate to twigs. His results, therefore, do not necessarily refute the camouflage hypothesis. The literature contains other examples of work on searching images, mostly inspired by the painstaking ecological studies of

L. Tinbergen (1950). These will be discussed later in relation to the evidence for apostatic selection.

(b) Searching images and colour polymorphism Apart from the recent work of Arnold (unpublished), to be discussed later, virtually nothing is known about the behaviour of çp-hunting thrushes. Certainly there is no direct evidence that they hunt this snail by morph-specific searching images, let alone preferentially search for the most common morph, as is required by the theory of apostatic selection. The same argument can be applied to all other non-mimetic polymorphisms. Yet some supporting evidence for apostasy does exist. This will be examined in the next Chapter. - 29 -

CHAPTER II, THE EVIDENCE FOR APOSTATIC SELECTION - 30 -

CHAPTER II THE EVIDENCE FOR APOSTATIC SELECTION

1. INTRODUCTION

So far, attention has been concentrated on the problem of how

colour-pattern pàlymorphisms are maintained. It has been pointed

out that the visible distinctness of different morphs: is suggestive of selection by predators hunting by morphw.specific searching images

This characteristic carries little iYeight as evidence for apostatic

selection when considered n isolation. It is the purpose of this Chapter to assemble and discuss the more conclusive findings that support the theory.

There are two main lines of evidence. The first Is provided by considerations of the effect apostasis would have on two sympatric

species exhibiting near-identical, polymorphisms (Clarke 1962b)

The second approach is more direct and concerns observations of prsdator..prc3y relationships

2. MIXED COLONIES OF CEPAEA

A testable prediction can be made as to the effect of apoatatic selection on two Closely related polymorphic. proy. It is essential for this prediction to be valid that the two species should: a. possess

visibly similar polymorphisms, b. live. in the same habitat, and c. share the same predators. If they satisfy these criteria, it is quite liiy that predators would treat them both as the same Species. This would mean that effectively one polymorphism would be maintained by predators selecting apoatatically. - 31-

If a particular morph is at a high frequency in one species for a reason unrelated to its visible properties, for example - then, if apostatic selection is acting, the same morph in the second species would be at an advantage only when present at a low frequency. The polymorphism would be brought to equilibrium at a point determined by the relative visibilities of the morphs against the background. Uniform backgrounds may produce an

'apostatic equilibrium point' with a high proportion of one phenotype in both apcio. In general, thongh,'9a series of visually similar but ecologically diverse habitats should exhibit negative relations between the frequencies of equivalent forms in the two species" (Clarke 1962b). In the absence of apostasy, postive relations would be expected.

Strangely, the only species to have been investigated with regard to the above prediction are Cepasa nemor'alia and Cepaoa hortensia (Clarke 1962b, Carter 1967, Clarke 1969). A glance at the literature will rsveal that the existence of negative relations is a matter of soma controversy • Their presence eisa apparently revealed by a sequence of samples from 'open' habitats near Oxford

(Clarke 1962b). Hors there was a statistically significant negative correlation between the frequencies of yellow effectively unbandod shells in the two species. In woods there was also a negative correlation, though non-significant. Likewise, Carter (1967) has found similar negative relationships within habitat classes, but has questioned whether they are duo to apostatic selection • In addition he has criticised Clarke's original data on a number of grounds, but without providing an alternative explanation for the negative correlations • A reply to Carter and a summary of the evidence for - 32 -

apoatatic selection are to be found in Clarke's (1969) paper.

Space does not aUot a fuller discussion of the Clarke - Carter'

controversy, but it is evident that many more collections of mixed

çpasa colonies should be made before pronouncing the final verdict.

3. PREDATOR-PREY STUDIES

(a) Introducing the evidence

Until recently there has been a lack of direct research on

apostatic selection and consequently much of its earlier support

derives from other lines of enquiry. The evidence from Reighard

(1908), Pepham(1941, 1942) and L. Tinbergen (1960) are cases in

point. Primarily as a result of the last author, new impetus was given to research by. ethologiots (e.g. Cross 1967, 1970; Wi. Tinbergon

ot al. 1967) interested in the behavioural aspects of predation and by population geneticists (e.g. Clarke 1962a, b; Arnold, unpublished) concerned with the implications of such behaviour to polymorphic prey.

Despite the diversity of its origins, the evidence falls into two groups • The first concerns experiments whore predators feed solely on one particular prey before a choice situation • In the second, such 'training' is lacking. The distinction between these two designs is important, for' though the first may reveal a mechanism

for apostasy, whereby predators become accustomed to searching for one type of prey, the second compares more realistically with the natural situation. Here, predators have the chance of becoming

conditioned to the first prey-types they encounter. According to the hypothesis of apotatic selection, they should on average become trained to the commoner types. - 33 -

(b)

Associative learning is a ueU.idocuinonted form of behaviour

(see Thorpe 1956, Chap. IV). Predators, for example, rapidly learn to associate specific colour-patterns with distasteful prey and hence become conditioned to avoiding them. (See c .g. Blest 1957, Sexton 1959, Brewer et al. 1960, Duncan and Sheppard 1964). The evolutionary importance of this typo of behaviour cannot be disputed for it has led to the development of warning-coloration and mimicry in my unrelated species. Is the converse of this behaviour also true; can predators become conditioned to search in for Dalatable prey of a specific- colour pattern?

(i) General evidenc

The classic work of Luuk Tinbergen (1960) provides an ,affirmative answer to this question. He shoved, in a comprehensive ecological survey of titnlce (Parus spp.), that birds can become so used to searching for a particular prey that later-emerging species are overlooked for a considerable time.. All these prey were Insect larvae, mainly of the Orders tepidoptora and Hymenoptera, inhabiting pine trees. A variety of techniques, including direct observation of food brought by parent tits to their young, enabled Tinber'gen to analyse the yearly factors determining the degree of larval predation.

Although the density of the caterpillars was a principle component, their size, palatability and conspicuousnoss were also involved. By taking these into account he was able to study the GffectS of pr density alone. - 3I -

When, for oample larvae of Acantholyda nemoralis and Panolia app. first appeared in the tits' hunting environment, they constituted only a very small proportion of the food brought to the young. During the next few days their densities rapidly increased and than remained constant. Theoretically the tits should have quickly exploited such concentrations of larvae, particularly as these species were known to be highly palatable. Instead, there was a lag of aboui a week, on average, before the birds started to exploit the new food supply, whereupon the larvae suddenly appeared in the nestlings' diet. Tinbergon decided that these results could best be explained by assuming that the tits were hunting by searching image. When the larvae first appeared in the environment they were taken entirely by chance and hence infrequently; the tits wore still soarching for prey they were more familiar with. As the tits found nore Acantholyda and Panolie, they were rewarded and gradually built up searching images specific to these prey. Total acquisition of these specific searching images led to the rapid increase in predation of the novel larvae.

Similar evidence for 'ecological' searching images has been provided by Nook at al. (1960) and Cibb (1962) • All these studies contend that the formation of a searching image to a specific insect prey leads to an increase In its risk of predation.. The important point to note, however, is that until this time the prey enjoys a certain amount of protection. If such behaviour is widespread among predators, it should be repeatable under laboratory condilfsne. The earliest experimental evidence comes from Roighard (1908), working with the grey snapper Lutlanust griceus (L.) as the predator and dyed sardines Athorina laticops CL.) as pray. His aim was to toot ci now disproved hypothesis (see Lorenz 1962) for coral-fish coloration. - 35 -

One hundred and fifty eoa..caught snappers in e large outdoor

pool were fad on Shout eighty sardines dyed blue. They wore then given a choice of five blue and five unfamiliar (either normal, red,

green or yellow) sardines. During presentation of the two colour.

types Reighaxd noted the order in which the sardines were eaten • Shen

blue and normal fish were offered, the latter type was understandably preferred. The remaining three tests revealed an overall preference

for blue. Apparently the snappers had acquired searching images for blue-dyed sardines • Even in the face of a multitude of predators

inhabiting an open pool e the other colour types were at an advantage, albeit short-lived.

Unfortunately, Roighard 'a apparent evidence for searching images may be deceptive. As Craze '(1967) points out, blue was the only colour-type used for training the snappers. It is wrong to ignore the possibility that blue was a colour preferred as a result of learning unrelated to the training.

In a comparable experiment, Smith (1967) attempted to train individual minnows, Phoxinue phoninus L., on one type of artificial prey (either white or blaàk worm-M6 pieces of a mixture of lard and flour). One group of five fish ww fed solely on the dark type whilst five other fish were given light prey. For three successive days, each minnow was given a thirty minute 'feeding period' on twenty monomorphic prey, scattered randomly on the sandy floor (2 That * 1 foot) of an aquarium. Thirty-six hours after the lest training period, each fish was presented with a 'population' of ten light and ton dark prey distributed at random within the tank.

Subsequent observations of predation revealed that, on average, the minnows significantly favoured taking the familiar colour. - 36 -

Similar results have been obtained by Bukeija (1968) who has rigorously investigated the increases in risk of novel prey types when

introducod to nine individual sticklebacks Casteroateus aculeatue I.

in an experimental tank wdor controlled conditions • The finding most relevant to the present discussion is that when the fish were

accustomed to feeding on Tubifex worc&, larvae of Drosophila were overlooked in preference to the former.

The experiments described above load weight to the argument

that predators continue searching for familiar prey when a novel

typo is introduced into the area of search • Such conditioning has been taken to an entreme by Rabinowitch (1961 9 1968). In one

experiment (1968) 9 chicks of captive gulls LXUS argentatue and Larus dolawaransie, wore fed on one of three restricted diets • These were.* chopped earthworms, pink Cat-food and green cat-food. After

five days of training each individual bird was offered a choice between first, the familiar food and one of the others, and second, between the two unfa!njljar foods • In both species there was a very

significant preference for the familiar diet. when confronted solely with unfamiliar food, some of the chicks totally refused to oat.

Comparable results have been obtained by Capretta (1969) for chicko of Callus gaUus.

Rabinowitch's earlier (1961) experiments were even more dramatic. Sin domestic chicks were reared on 'silo' from hatching and eight were similarly brought up on whoat • After sin weeks the dieto were revereed • So conditioned were they that two of the silo-trained and five of the wheat-trained chicks refused to oat the unfamiliar food and actually starved to death. - 37 -

These results with young birds agree with L. Tinbergen's (1960) contention that a prey is not recognised as food until tts reinforcing value has been established through ozperienco. In this case, however, the avoidance of unfamiliar food was probably OxaWarated as a result of the predators being at a highly impressionable ago. A searching image should normally be more flexible, for it is to the predator's advantage if it cn quickly switch attention to a particular prey. The main stimulus involved in searching image behaviour appears to be the Colour (or tone) of the prey • Dawkins (1966) has shown

that domestic chicks will, on aquiring a searching Imagej hunt for prey on the basin of colour alone • Cross (1967) has demonstrated a similar effect in carrion crows, Corvus corona I. For example, when a white cockle covering a piece of meat was presented to a crow

('Abendogo') on a beach, it subsequently turned over two white extraneous shells, a piece of white fluff, a broken white mussel shell and a dead rabbit in addition to seven white cookies, of the experiment proper, and a black atone • The latter incident was presumably an example of a chance encounter with an unfamiliar object (see page 4-). In another study Craze (pcit.) enquired more deeply into the apocificity' of searching images. Crows were trained to Overturn mussel shells painted red and dintz'ithtod on a beach. In the choice situation, those were presented in equal numbers with another unfamiliar form. They were distinguishable from the standard shells by one of three criteria: colour, shape or surface structure. Generally, his results were that the chances of the unfamiliar Lore being taken depended on the degree of colour resemblance between it and the standard rod shell. Thus ro&yoliow mussels were overturned as often as the - 38 -

standard rods, tyhilot yol].o'a russe10 tere totally overlooked. In comparison with colour, shape and surface touture were loss Important visual cues; red cockles, for oxasplo, were treated on equal terms with rod mussels. Until now, we have discussed the evidence only in relation to birds and fish, animals that mainly hunt by the sense of eight. It has recently been demonstrated that olfactivo cues can also elicit aerching images in prodatorm • Soana (1970) has boon able to condition captive rodents (pdemus sylvaticus L. and Mus ausculu Ia.) to search for artificial prey with a specific scant. During choice situations the familiar type was taken in preference to prey that had a different scene but identical appearance • Soano points out that there is scant literature on animal scent polymorphisms (but see 11er 1878 for scant dimorphism in the males of the butterfly, Bttus polydamus). This situation may simply reflect lack of research. If future work reveals a wealth of animal species polymorphic for scent, then we must not overlook the possibility that such variation is being maintained by apoatetic selection. The main implication to be derived from all tha findings discussed above is that predators may be capable of becoming conditioned to specific variants of a natural polymorphic animal. That this occurs, is, as will be seen, far from conclusive.

(ii) Natural ôlymophic pre The only polymorphic species that has been investigated in this context is Cepaoa hortonsis (Clarke 1961)1.

Two pairs of hand-reared aongthruehea were housed in separate sixteen -foot square cages • Each cage was divided into a straw-lined

1Since the time of writing * the work of Don Boor (1971) has been published (Sec page 265 ). - 39 -

'collecting-area' and a 'thruh'.stone area'. Snails presented to the birds in the collecting area were taken to the thrush-stone site for breaking. Here the broken shells could be scored and counted. The experiment involved only two phenotypes, y0ilou unbanded (Y00000) and yellow banded (Y12345) morpho. Possible errors from differences in taste and behaviour were minimised. For three days the birds were trained on yellow unbendedin one cage (cage I) and yellow bandeds in the other (cage IX). They wore then presented with 'populations' of the two morphs in unequal proportions: 260 ?00000 's and 40 712 3L5 'e in cage I

99 and &O " 260 ' " cage XI.

On each day, the btokeu shells were replaced by the same uuthev of living snails with identical proportions of the two morphG. Hence the composition of the population was kept constan1. The experiment was terminated when about 700 snails had been eaten. It was then repeated, the only difference being that the straw was slightly darker than before.

Clarke found no clear evidence of conditioning, since banded forms were preferred in every case, oven when the birds had previously been trained on unbandeds. In only one instance was the preference statistically significant; this was in the first experiment, cage I. When the results for both experiments in cage II were added together, banded forms were again significantly favoured. A possible factor contributing to these results is that the thrushes may have become very efficient at searching for Cepaca (cf. do Ruiter 1952) since this activity was virtually continuous (Clarke, personal communication). For instance, the shells may have been detected on the basis of tactile cues. The experiments have never been repeated. - -

- 100 ACANTHOLYDA •--:

0 0 o 10 U.

li.. 0 •

I.. z ic hi U - • a hi CL

01 (Z.______- oo-o oe • • 001 0.1 10 DENSITY

The rltionship between the density of A=nth2l Zdg ooli8 md its proportion: in the tits tot. otQ that at low dnitioo tho lamm wero taken boo ofton then oupocted (continuous lino) cd at high donsitieo more ofton • From Claxko (1962o); hie Fig.7. - 41 -

We can conclude from the preceding discussion that predators may sometimes acquire searching images after several encounters with a particular type of palatable prey. If this prey is the commonest variety of a polymorphic species then selection will be apostatic. On the other hand, a rare morph may be the first type to be encountered by a naive predator. If the predator finds it unprofitable to hunt by using a searching image specific to this morph, it may possibly switch to searching for a commoner variety. Unless it does so, it will not select apostatically.

We must now examine the literature for evidence as to whether predatorsdo, on average, select in an apostatic manner.

(I) General evidence

Reference is again made to the work of Tinbergen (1960). Our interest centres on the situation when the tits were feeding on a number of larvae simultaneously. Tinbergen found that at low densities certain caterpillars were taken less often than expected. At moderate densities expectation was exceeded, whilst at very high densities tits prefer a varied to a monotonous diet and so dispense with searching images. Fig. 2 illustrates these findings, with regard to Acantholyda nemoralis.

Strictly, these data cannot be taken as evidence for apostasis, since Tinbergen was dealing with changes in density, not frequency (Croze 1967, 1970), - 42 -

For genuine apoetatic selection, as the density of Acantholyda. increased, so the density of alternative prey should have decreased proportionally. This in fact did not always happen. r. J.N.t1. Smith has drawn my attention to an additional criticioc of Tinborgen 'o work (including the experiments mentioned on p.33). Predators probably deliberately seek out areas of local high prey density (see, for example, Goso-Cuotard's 1970 data for feeding by the redshsk Trip totanuc3 L.) • If the tito had acquired searching images for areas of high prey density (as oppood to for opacific typos of prey), then Tinbergon 'o estimates for overall prey density may not have been representative of the areas visited by the birds. However, the work of Halting (1965) has allo demonstrated similar responses to increases in prey density. He used captive small maiinals, including deer mice. Peroocue leucopua Rafines quo, as predators and cocoons of the sawfly Neodriprion sertifer Geoff. as prey • 7hon alternative food was available, the nuober, of cocoons taken by the predators varied with density in a manner similar

to the response of tits to Aeanthol24 as shown in Fig. 2. More otperimontal evidonco comes from the work of Popham (19 141). He investigated the selective value of crypois in the corixid bug Arctocoriea (Sigara) distincta L • Adults of these hemipterans can be various tones of brown, depending on the shade of the

background on which the larvae develop. In one set of experiments two varieties were offered to tudd, Scardinius oz opbtha3ius L. One of the morphs matched the underlying sand, whilst the other was non-cryptic. In each experiment different proportions of the tx) varieties tiers presented to the fish and those ratios were kept 43 Pimutoo aa and

Cicio ü ramalyola o2 tho data fton ?hcn (191 and 1932). The Figwo end Locade ao takan ftvm C1azco 1962ci (Ttrj. Se ic O quivOIGUT to C3.ao'o F11. S md Tie. 2) comoopmda with hic Fits. 6).

100 • 90

30 V V / I70 '0 60 so

VV V • 40 V

- 30

• V 2

V • - 0 10 20 30 40 50 60 70 80 90 100

V PERCENT. OFFERED -

__3. The rosulto of pMdation by dd Scwdlniun. (LOUCIOCtn- on oidd bwjo jai otinet. n c' ooIoa oopozL t to c1 u..ty9oo of buo woro ood to the prodator in di3t proportlono. The bQek 4 Tiao og uali7oto colcur and too ('i' on the 001d Colo chat). The horiMatal MIG iivoe the popotion of an y eo1otypo ofovd, the vart ical cdo ohoo the tion Vkon by the r:dd. 0on ofreloc r000nt o1ouztyo 'i (czpptic) and csood ct1oo mpr000nt colour typo °1,° (akor md loco yptic). Each aoitnont, thIh vaDrasonto about 200 °podationo' Io thoveoo ropmeontod by tvo cic1o3, ono open ond one wwood. For ca anp1mation of the ctioio md dattod liziioo, oeo to1 • Date frozn Popba (11).

90 / VV/

- 00. V' / V 70 Q

•• 60 0,

Q' I I_so •e • V V • • V Z 40 / V • • , I .

V - • 20 • V

• • V I , /4 00

- 102030405060708090 1 00

V PERCENT. OFFERED

Fiçuo 2. A diaaz ek11a to M g . 38. givInc the ra sulto 09 omhor oozieo ©g oxpipl=nta it th$.h cm additional ce1tour.tyo voo ictoducod

(001otyo InO, ovon dav~tev E4 1iQo cptio thm 11 1 ) .Thia ic reproscatod by the b1th cizcloo. Noto the ch maa g in aU colour-types,, froo advantao. oco at 10 onoioe to dLoc6vcmtn3oowneoo at bL onoo • Data fror payhm (1042). - 44 -

Constant by replacing oach eaten bug with another of the same colour.

Pophazn'o data have been z'eenalysod by Clarke (1962a). The.

results of tho first experiments are presented graphically in rig. 3a.

Clearly, the non-cryptic variety was over.prsdated at all frequencies. Using the data obtained from when the phases were presented in equal proportions, Clarke calculated the expected proportions taken from experiments involving unequal numbers of the varieties • These expected values are shown by the continuous lines in Fig. 3a. We can now see that the disadvantage of the non-cryptic morph decreased at low frouoncioo and incroasod at high. This effect was statistically significant. In a second series of experiments (Popham 1942) when another variety of corizid was added to the system, there was again a

definite advantage for a variety at a low frequency. Correspondingly, each typo became significantly disadvantageous when relatively more

abundant (see Fig. 3b). Hera appears to be clear evidence for apootasis. However the results wore obtained from only three predators, and Popham himself showed that cox'ixld bugs become restless when on a 1xatk -gmQ=,d of a colour differing from their own • These two factors may have imparted error to the experiments.

Smith (1967) carried out a sot of experiments along the sane lines as those of Popharu. In this case, minnows Phoxinus phoxinue L. were presented with populations of dark or light artificial prey.

He used large numbers of fish and offered different proportions of pray to each individual. Under such conditions, he was unable to duplicate Popham's results • This was apparently because the fish were so variable in behaviour; any apostatio component of selection - 45 -

100 0

0•

41 80-

LU A 0 0 0 60 oc 0, I-z , 0 40 , LU 0 A CL , , 20

0 , /

C, 60 80 100 0 A 20 40

PERCENT OFFERED

£ The reeulte of predation by minuous (Pho,dnuB hoxinus) on dimorphic populations of artificial prey • Each trial ie represented by two symbols an open symbol desiating the proportions of light prey offered and taken and a filled-in symbol denoting the percentage of dark prey offered and taken • The circles and triangles designate trials on two different backgrounds. pots the large variation in results compared with the expected valo (dotted line). From Smith (1967). - 46 -

would have been ccmplotely masked (see Fig.

Individual differences in predator behaviour have also boon shown

by Pough (1964. In an attempt to test apestatic oclection, be offered rod and yellow painted sunflower osode, in ratios of 9:1, to caged chickadees (Parus a. atricapiUus). However, the: heterogeneity in behaviour was not enough to conceal overall trends, for each bird tended to prefer the colour at the lower frequency. That is to say s the predators were behaving in a manner completely opposite to that predicted by epostatic selection. The prey were probably at too high a density.

This point will be discussed in more detail later (coo Chapter IV).

(U) Natural plvor1)hLc pr2

We are again forced to turn to Smaea the only opeciec that seems to have been satifactori1y investigated in the present context. Reference is made to some unpublished experiments carried out by Dr. R.W. Arnold (personal communication). He sot up artificial colonies

of .Cel2aea nernoralis at different sites in beech woods • These populatins each contained over 1000 snails and differed from one another with regard to the relative frequencies of the morphs that they contained. Only two varieties were prosentedt unbanded yellows and unbanded browns. Thrushes regularly visited the qopulations and their anvils were periodically visited in order to assess the predation. Yellows were always taken to excess • That is, relatively too many yellows were taken from populations where this colour was presented at low frequencies as well as from populations were yellows were cownon • But by taking this preference into account, Arnold was able to show that selection was in -47- fact frequency-dependent.

4 CONCLUSIONS

From the preceding review of the evidence, three main conclusions can be made with regard to:

conditioning to a specific prey; the specificity of conditioning;

apostatic selection.

conditioning to a specific prey Sight-dependent predators can, under certain conditions, become accustomed to searching for a prey of a particular colour or pattern. The frequency with which this occurs in nature is unknown; few of the experiments described concerned predators in their every-day environments. Of these studies, the evidence from Tirthergen (1960), Mook et al. (1960), Gibb (1962) are least reliable since selection may have been based on other cues besides the colour or pattern of the insect prey. The most satisfactory data are those from Croze (1967, 1970).

The specificity of conditioning Apart from the work of Croze (1967, 1970) little is known about the specificity of this conditioning. Yt this is of crucial importance, since the theory of apostatic selection assumes that predators acquire morph-specific searching images. Attempts at conditioning prddators to a particular morph of Caea have failed

(Clarke 1961), though the occurrence of such conditioning is implied from the results of Arnold (unpublished) and from Clarket -s (1962b) work on mixed colonies of pea species. - 48 - - -

Even if eight-dependent predators do, in fact, become conditioned to searching for a particular variety of a polymorphic prey, there to no logic in assuming that they select apoetatically. Yet it is critical to the theory that, overall, they should hunt for the commoner vax'ietles.. Arnold (unpublished) has provided the only aatisfaotory demonstration that predators (individually or overall) select prey in a freque cy"depondent manner.

The not Chapter discusses the basic techniques that gore used in experiments designed to investigate the above and other related topics. - '49 -

CHAPTER XI! EERAL NATERIALS AND tZTHODS CHAPTER III GENERAL MATERIALS AND METHODS

o blackbird! sing me something well: Uhile all the neighbours shoot thee round, I keep smooth plots of fruitful ground thoro thou !aayst warble, eat and dwell.

(Tennyson 18e2).

1. INTRODUCTION

This research was specifically aimod at testing apoetatic selection,, and in this respect differs from most of the work discussed in the previous Chapter. Tho approach to the problem was basically simple: wild birds in their natural surroundings wore offered populations of polymorphic artificial prey.

2 • CHOICE OF EXPERIMENTAL SYSTEM

It must be remembered that oxporiontal conditions, especially in the case of animals in captivity, may distort responses to the: prey • In the present experiments the birds were in their usual habitats (gardens or fields) and were therefore sub]oct to environmental

Influences outwith the experiments per so. In this respect the method differs from those involving caged birds and is nearer the natural situation. Ideally, predators should be directly observed feeding on polymorphic prey in the field. Unfortunately, accurate observation and identification are difficult under such conditions. Predators are active creatures; a meadow-pipit feeding on PhLlaenus in long. grass, for example, would probably capture and swallow its prey almost instantaneously. Even assuming he saw this act an Observer would have great difficulty in iiemtifying the species of prey caught, lot alone its variety.

An improvement on this approach is to present predators with - 51 -

artificial populations of their natural polymorphic prey. In this way the site of the ouporimental population can be controlled and can be so designed as to be fully observable to the researcher. Problems of practicality ar's, hotievor, still involved. An important potential source of error concerns the possible relative behaviour (e.g. mobility) of different mor'pha • This night be ovmrcoma by killing the animals before presentation, but most tdld predator's do not normally eat their usual prey that are already dead. Katz (1937) tells of frogs that starved by the side of a heap of dead flies; they did not rocog Leo the motionless insects as edible. In zoos, penguins are usually fed on dead fish. During eighteen year's of observations on king penguins (ptenodytec atgonica) Gilloopio (1932) did not note a single case of them taking a fish from the ground. At feeding time over bird had to be fed singly by hand. It was decided, therefore, to employ artificial polymorphic prey. Those are immobile and 'novel' and have the added advantage that they can be made desirable to many species of birds • Moreover they overcome the problem of obtaining large numbers of spOcified phenotypes. (round-fce6ing passerines are ideal for prey-selection experiments. Apart from their abundance and ubiquity, their main assets are that they hunt almost exclusively by sight and possess colour vision (Puinphy 1948 9 Kettlete11 1955) and are extremely efficient at searching for prey (of. Hoppnox' 1965, for Turdus mirator'ius).

Many psssoriner3 undoubtedly food on polymorphic prey in the wild.

3. BIRDS

The species of birds involved are some of those normally to be

found feeding on lawns or fields. A full list if given in Table 3.

Further details of feeding habits and general behaviour can be found

In Uitbcry at al. (1938). (i Table :1

List of birds involved in experiments

[Ciwnon name c name

Blackbird Tur'dua mezula L.

Songthruab Turdus philontelos Hart (z T. ericatorum Turtan.)

I Robin Er.thacus rbecula Hat. Dunnock (hedge sparrow) Prune lie noda1arie Hirt.

House sparrow Passer . dconesticus L.

Starling Stua_u1aris. L. Blue tit !arus caez'ulous L.

Table 2, Colours of the green and brown baits based on the Nunsell co1oin System (1966)

Value Hue wona 7 Green I 7.3 GY 10

Brem I 5 YR

Note. The hue of a colour indicatso its relation to red, yellow, green, blue and purple; the value its relative lightness or gray value; and the chrosa, its strength. - 53 -

The cost important predator was the blackbird (for Ito hebits1 Gee Snow 1958). This was always tho ft, and cornetimos tho only, species to feed on the oporim3ntal populationo. In genera1 additional species participated in otperiisnt@ carried out on lama or in market gardens,, in urban end ouburben areas • On the other hands all the studies perforeod in rural fields involved blQClthirdi3 alone.

. PREY

In ncst experiments* the prey wore groan or brown cylindrical pollets t 0.7 an long X 0.7 an wide and a description of their appearance and manufacture now follows • Other typos of prey will be discussed in the relevant Chapters (IV and IX).

(a) NoMh_olo mt The marphe differed only in colour. Plate 1 to come oxtont

Illustrates this contrast, whilst Table 2 lists the difforencoc more- precisely by giving the scores of the two colours according to the 1mcoll Colour Sytot (1986). Green and brown wore chosen for the

original oxperlm3nts (Allen 1967,and see Chapters V and VI) because it was believed that they would bo representative of natural palatable

foods • The results of these eporioents dictated that the easo. colours should be employed in most of the ensuing work. 1th regard to other properties the morphs wore apparently identital. The baits were of a manageable else for both experimenter

and blrdei from my point of view they could be easily handled, and most passerines could swallow thorn whole • Care was taken to ensure

that the prey (or 'baits') were made of approximately equal size. - 34 -

Plate 1

2 standard brown and 2 standard green baits on a grass lawn • Each bait is 0.7 an (approx. a") in length. - 55 -

There was no evidence for large differences in taste (see Chapter IV).

(b.) Manufacture

The prey were made from plain flour ('Snowy Mountain') and lard ('White Cap') in a 5 to 2 ratio by weight (after Turner 1961).

Kneading produced a malleable dough of 'plasticine'-like consistency.

Artificial food dyes (varieties 'green' and 'brown', supplied by the Scottish Colour and Flavour Company) were then added, 11cc of dye per 1000 gins of dough giving the required colours.

Once the kneaded mass of dough had attained a uniform coloration, it was forced by hand through holes (0.7 cm diameter) drilled in a sheet of perspex. The long 'worms' so-formed were laid side-by-side and simultaneously chopped into 0.7 cm-long sections. Large numbers of baits were made at a time and were stored in plastic boxes at room temperature until required.

5. ENVIRONMENTS As already mentioned, the experiments took place in gardens or fields. The background was usually grass, though in some cases soil was used (see Chapters V and VI). To the human eye, browns appeared slightly more conspicuous against grass, while greens were slightly more conspicuous against soil (pp. 118-119). Rain dulled the colours and made the baits soft and mushy, causing them to disintegrate, so that they had to be frequently replaced. There was no evidence that wet baits were treated any differently from dry ones. Experiments at different sites had different environmental variables and these were often correlated with climatic conditions. The birds (with regard to both species and individuals) were not necessarily the same at any two or more given sites and, in general, predators may be more selective in summer than in winter (cf.Prop 1960). The backgrounc at different sites also differed; the well-hewn grass of a summer lawn - 56 - is a more uniform substrate than, say, a winter field.

Variation also occurred within experiments. There was no guarantee that the composition of predators was the same at any two points of time within a given experiment (e.g. see p. 96), unless the birds were colour-ringed (e.g. see Chapter VII). Fluctuations in background conditions were also liable to occur, particularly in summer, because of mowing, rain or drought. The last two variables affected the overall colour of grass, but the differences within each experiment were slight. Rain only prevented birds from feeding when it was very heavy. Snow lay on the ground during parts of some of the experiments described in Chapter VI, but the effect of this variable was remarkably slight (see pp. 118-119). In other words, the bait populations had many variables in common with natural prey populations. We shall return to some of them at later stages. Despite the environmental differences between experiments, the results were often strikingly similar.

6. TECHNIQUES

(a) Presentation of prey At some stage in each experiment a population of randomly distributed prey was presented to the birds. The baits were usually at a density of two per square metre (for the exceptions see Chapter

IV) and the procedure of establishing one such population was as follows: (i) Experimental ]?lot First, the area to contain the population was marked out. This comprised a rectangular or square grid of one-metre squares delineated by metal or wooden skewers • Each column and row was respectively numbere and lettered by upright skewered cards around the perimeter and, in Figure 5 'Roneo'-copy of grid used on 22.e.69 in ' L c 9 9 Io experiment 5.5, Chapter VI • showing the distribution of 180 greens and 20 browns. The squares with double lines represent itiinaiiiIulial the one-metre squares of the plot. Own -- Skewers were placed at points equivalent Uivaii IMOOMENIMMIMID to the intersections of the do1&.1nes. 3 The four quadrants of each metre-square are also shown • For details of the randomization of the spatial distribution, C see text. 5LORIMMOMPUM1001001501 1OnEyMMOUBMIUM1001001 E SNEED MMUNIMMUMEMI F MIM-100 On M! ii DD 1! 11111 - 58

some cases, along the middle lines • This facilitated identification

of each metro-square. Physical and practical. factors limited the size

of tho plot but generally it was 100 square metres, corresponding to £1 population also of 200 baits.

(ii) Randomization of py distribution

The randomness of the distribution of the prey was an important component of the experimental system and the methods by which it was attained deserve full discusio. The first stage was to make

'Ronoo'..copios of the grid (Fig.5) with oath metre-square divided into four quadrants. These copies were then used to record the distributions of the future populations. Randomization of a given population was carried out in two stages, as follows: First, the baits were randomized with respect to position. The four quadrants of each motro..square wore labelled clockwise from 1 to 4 and then from 5 to 8 • Thus each quadrant was represented by two numbers. Tables of random number's (Fisher and Yates 1963) were now employed. Different rows or columns of these tables were used for each distribution. As the random numbers were reed, the sequence of occurrence of the numbers 1 to 8 gave the spatial distribution of the baits. The following points were observed:

A • Tho copy of the grid was filled in row by row. B • The positions (with respect to quadrants) of two baits were marked per metro-square.

C • Theo two baits were never positioned within the same quadrant. This necessitated the rejection of random numbers that followed one designating the same quadrant.

D • Random numbers 0 and 9 wore ignored.

Once the future poitiénn of the baits had been established,, their - 59 -

distribution with respect of colour was computed. In the early

experiments random numbers were used. For instance when two colours wore to be presented in equal proportions (e.g. see Chapter V), half

the random numbers (0.-5) designated one colour, say green, and the

rest (6-9) represented browns • The colours of the baits were then

superimposed on their positions according to the Order of the random numbers that were road. However this procedure frequently resulted in

discrepancies from the required proportions, and the colours of some of the baits therefore had to be changed. This process required further randomization

In order to quicken the operation later experiments Involved the use of playing cards • Thus in 1 1:1' experiments, equal proportions of cards of red and black suits were oiaployod. One colour represented green and the other brown • Care was taken to ensure that the cards were thoroughly shuffled before use. They wore dealt one by one and the ensuing order of the colours was systematically superimposed on their positions shown in the plan.

During the preliminary experiments it wad discovered that the plots may not be searched uniformly by the birds; for instance, they may prefer one particular corner. These birds would be exposed to the morph-frequencies only within this particular area and such frequencies might differ from the prportions within the population as a whole.'

In order to combat this possible source of error, the grids of later experiments were divided into four equal areas • Each of these sections Contained identical proportions of the morphs. Thus in the case of populations with equal numbers of greens and browns, each of the four areas had baits in ratios of 1:1 • This procedure mm imi sod the possibility of large differences in morph frequencies between different parts of the population.

Using the plan as a guide, the prey were now spread within the - 60 - grid. To eliminate possible bias in the positioning of baits relative to ground cover, they were dropped into their appropriate quadrants from waist height. Further randomization was added to the system by frequent changing of the distribution. This was usually carried out daily and ensured that the birds were unable to learn the positions of the baits.

(b) Recording of data Each population was observed from suitable cover; for example, a room or a 'Land-Rover'. Binoculars were essential for watching the behaviour of the birds • When a bird entered the plot, the events were chronicled in the following order: (1) Time of entry. Species of bird involved and, if possible, its sex, the colour of its ring (see Chapters VII, VIII and IX), and other recognition marks, Positions of baits 'eaten' or 'imprint-pecked'. This was done by tracing the paths of each bird on to the plan and thus marking the positions' of the relevant prey. If all or part of a bait was eaten it was defined as 'eaten'. But if it was picked up and then dropped it was defined as 'imprint-pecked' (see Chapters VII and VIII) and 'uneaten';

The colour sequences of baits eaten and imprint-pecked

(actually recorded under (iii)).

Other points of interest, for example behavioural aspects relevant to feeding.

Time of departure. This was rarely possible to record accurately when more than one bird was feeding.

With practice, two or more birds could be observed simultaneously but when predation became too heavy only a few selected individuals could be followed. The overall proportion taken by the rest was - 61 -

deduced by subtracting the numbers of Observed eaten baits from the total numbers actually missing.

All experimental populations wore at some Stage observed in the above manner, but some (see Chapter VI) wore'also subjected to 'ueohservd' predation, This simply moens that they were left unattended for o few hours and then ozaminod on return, in order to determine the proportions taken.

7 • PLAN OF EXPERIMENTS

The campaign to test apôstatic selection evolved along two main pathways. The first involved determining whether birds can

boapTaS conditioned to search for a particular colour (see Chapter XI, pp.33 .39). Ujid passerines were fed on either greens or browns

end wore then presented with a mixture of both In equal proportions. After a proliminaty experiment, a sorioo of duplicates was carried

out. Later experiments were designed to test the specificity of

conditioning. Those last studies involved the presentation of additional colours of various shades of khaki.

The second approach was to discover whether untrained birds. preferentially search for the commoner colour in a mixture of two morphs (see Chapter II., pp.40-. 47). Populations of greens and browns, with one colour nine times as common as the other, were presented, These '9:1' experiments were repeated a number of times and zosulto were obtained for selection by both individual birds and groups of birds.

Finally, the two strategies converged on experiments involving the presentation of 'normally distributed' polymorphic populations containing greens, browns and various shades of khaki. Piptire 6. Interrelations of Chapters IV to

LOW DENSITY EXPERIMNTS

EXPTS WITH GREENS, PRELIMINARY EXPTS. DUPLICATE EXPTS. BROINS & INTEPJ'IEDIATE WITH GREENS & BROWNS. WITH GREENS & BROWNS COLOURS

1QjI1,1M1 DENSITY EXPERIMENTS . 1:1 popns. after ' 1:1 popns. after 1:1 popns. after training . training training on khakis

UCUS JOURS OP GREENS & BROWNS V V '-•' IALL' BAITS 4' Polymorphic 'normally distributed' popns. With & without previous training

9:1 popns0 Responses Introduction of a of individual birds 3rd (khaki) morph (Thrdidae) into the population

9:1 popns. Responses VII 1 VIII IV IV IV of birds en masse

( S.

S . VI 9:1 popns. Responses of birds on masse

VI -63-

All these populations tier presented at densities of tso baits per square metre • In an additional sot of ocperimonts populations were offered at much higher densities. One of the aims of this study was to test'unifying' selection (Pieloiski 19S9) which In GfEact, is the opposite of apostatic selection.

Figure 6 summarises the evolution of the work and rcalites the experiments to each of the following five. Chapters. It should assist the reader to visualiso the connexions between the several major typos of operimonts • Of these, we first describe and discuss those that wore carried out at high, densities.

0 - 64 -

CHAPTER TV WAXX11U.DENSITY EXPERIMENTS - 65 -

CHAPTER TV 1AXflU'DENSITY ERI1ENTS

1, INTRODUCTION

This Chapter describes experiments with dense populations of baits • Consider the case of birds feeding from a bowl containing a mass of standard greens and browns in equal proportions. If predation departs significantly from the expected 121 ratio then it can be assumed that for some reason one morph is relatively more attractive than the other. Let us nci imagine that the colour's are presented in unequal proportions with, for example, greens nine times as common as browns. By taking the frequency..indopondant colour preferences into account we can dorivo values for the expected predation. If aposteitic selection acts at such high densities then the birds should take relatively more greens then expected. However, the literature contains evidence suggesting that the

opposite is the case: rex's types are likely to be picked out from a group. For example, Piolowaki (1959 and 1961) has shown that

goshawks (Acobiter geniti1is L.) take white pigeons more often when most of the flock is black, and black ones more often when most

of the flock is white • Similar results have been obtained for other predator-'prey systems by Tiuborgen (1946), Willard (1966, quoted by salt(1967j. and Pough (1964). The experiments carried out by the

last worker wore supposedly designed to test apostatic selection and

will be discussed later in the Chapter.

The main aim, then, was to discover just how wild birds respond

to populations containing baits, at vary high densities and in unequal Proportions. This line of inquiry progressed from using standard

Croons and browns in the first series of experiments to employing -

Nitr Mir

jit -, 0-- 1~ 74

01 ter

.7,':. .. •,,.. 4h •'1'b, tg4 - • 4 1 ' ••' ; V 4L ,.

Plate 2

Maximum-density experiments (series 1). Populati on 1a 150 standard greens + 150 standard browns randomly distributed within a metal sieve 19 cm in diameter.

... . .., •:.-, • - 'L-' - : - .'-' r• •''

. :-..

.. •

• .•'--_ .. • b C d - • ,,,, - frI - ç . .L. -aVi .... ...... '-.•i,. fr' +' • ,i.if

.-.: ...r ••.- .

-- 1 • - -, •.- . ilk .

Plate 3

Maximum- density experiments (series 2). 1:1 populations: an lift, 70 small greens + 70 small browns; on right, 70 small reds + 70 small yellow.. The p.tri - dishes are 9 cm in diameter. - 67 -

Oma1.to' baits of thoco Cmd other Colaum, in tho occnd

2 •UATRXA1S MD MTHODS

(G) Sôzioo 1: The pOpn1LtiGe OZ ocandad amon id bn baits wom poonted

In ci'1a (19 cz deto) MotOl Slovesom WrZ Agemont that onood that ttoz did not collect. In late Uarch 1968 9 150 gmon and 150 barn baitø randomly diota'ib*tod within a eiovo (Plato 2), thL& wee then placed an the oda of a 3m outaido the Dpattent of ZooloMr. ui7e2r3ity of Edthu. The hattc woo voy &.eeoly pedtoft in fact

they wora at the amimm p000lble dcnoity for eocte lying en a plane

otfaco. Podia1137 9 thic poplaticn (la) tieD oeDinod and, after

MCOT&ng the ntm-baro of baits takes e, ww roccnotittod to Ito oriSinal cio. i?oqut cboervotlow frorn thQ a4jolaing building dotenined that the solo podatozo woro blacktArdo, probably jmt a malo and his tate,

After 6 dare of pxwentation o, nd CM intovai of about two toekc,

the 131 populaticatiea ropMcad by ttio othor3 of the oao oio one with gono nine tioo nowo awiinon than bmmc (ib) nd the other with bwnr nine tinco., M eomm ae zeono (lc) The eioo ccntining tbieeo populatione tezo placed one Lmtro apart and tholp poeitiono relativa to one mother wore FmqUmtly and rmd=ly changed. In othav reopoctci the operitont siee cied out co bofio. Dias the oihth day, the blbirdo uovo joined by an i otoinzblo ntmibov of hou000 rwe and the to of Lovedation r=o ocodin1y. The onper4cont leotod ovo' a

(b) Soioo2 The odent of ym&otlaa by h©optio during the 1ao oiiant - 68 - intoduod a eowce of aor. These birde tended to pock the bcito ithe -tham oat thom ho.o and small places tcto coqent1y loft In the 0iove3 • Th000 wom roughly meanotitutod into whole boite In oz to dorlvo en ootiiato of the nmzbara of whole baito eoton. In the ozinto that onood, thie pz'oblora was ocozo by pmcantlna baits of ouch a oio that botmeoparrms could otiailow them ho1o.. Theoo baits aara romd (app. 0.3 cm dito) polloto thich wom made foa otiafl piceaz of co1ouod 1exd and flour rou1dod botoen thwch and feiethor.. In oitparltmnt 2a they wom gmen and ban o boftro r, Thiot in 2b they wore yal1w end rod (oe Tb10 3 end Plate 8).

Tthlo 3. Scod by the HeMS011 cO1Ou SYOUM 1966 (oee p.52) the rods and yol1cio had values as chcrn.

Value Cbici

ST 10

Rod 7.$RP 10

The popuXotioe of theoo baits wore proentod in cicu1e' (9 cra in dierto) pleetic potidihoa with small ho3io drilled in thofr be!ee for drainage. Each dish hold 140 baits and the diohoo waro p'eacnted in a etaibt line and placed 50 cm apart.. During the oMorl=nte the populatieno ioo maintained ae before. Epoziment 2e was ot0tdd on 10th Nay 1968 k whom 1:9 0 9:1 and 1l ppu1atianc trara offood to the bfrda • Afto four days, 7:1 and i0 populations woz' oddod and the opoicant centinted for four mom deyo. EzpozLeat 2b an 20th tey 1960 and 2611cod the oazo plan ac Ito podece000. Table 4

Grand totals of green and brown baits taken from populations

presented during experiment 1.

Proportions offered Dates Predation (1968) G B

la. 1G:1B 31.3 - 604 202 226 Expected 1:1 214 214

lb. 9G:1B 22.4 - 27.5 1606 592 Expected 9:1 1978.2 219.8

Expected constant selection 2027.7 170.3

ic. 1G:9B 22.4 - 27.5 743 942 Expected 1:9 168.5 1516,5

Expected constant selection 215.9 1469.1 - 70 -

3. RESULTS

(a) Series ii (1) Synopsis

Table 4 gives the grand totals of baits taken from the 1:l and

two 9:1 populations of standard greens and brona. The values for expected predation, assuming random selection, are also shown. The

results for the 1:1 experiment suggest a slight preference for brown

but the deviation from expected is not significant Q 2,1.3S, P> 0.20) • On the other hand, both the 9:1 populations suffered

heavy overpredation of the rarer morph'. Those deviations from expected are very highly significant (when brown was rax'e, X1) 700.309 P ' 0.001; when green rare, X2 ) 2176.80 0 P c 0.001). Clearly, the birds were exorcising selection against the rarer colours and this was of such a magnitude that it overede any selection due to frequency-independent colour pro ferancos.

Nonetheless, another type of selection is in fact, detectable. The two populations containing the colours in w%oqual proportions

were carried out. simultaneously and should have been subjected to identical environmental influences, including numbers of visits by

predators. It is valid, therefore, to compare directly the data

collected from experiments lb and Ia. If this is done, we find that as well as preferring the rarer baits, the birds tended to ta}to green ones • This tendency can be detected by comparing the proportions of

common and rare baits taken frOm each population. The null hypothesis is that these ratios are the same, assuming constant selection against the rare forms, Thin is not the case (x 1) 128.51,, P CO.001). It can be deduced from the data that this heterogeneity is due to the

'Some of thoso results iavé now been pubifehod (Allen 1972). - 7]. -

Table 5

Nba9 of baits taken by ipale (O') &ad female () blackbirds during cbsorvd th3ite in oxperiinant la.

bsze5 taken Data observation

C B 31.3.68 19.01 0 3 09

14.00 12 0 0"

18.0 13 0 0"

3.14.68 11.10 3 0

11.29 2 0

11.31 7 0

4.14.68 9.30 0 17 a" 9.31 3 0

6.4.68 9.29 1

Grand totc-JD 39 27 predation of relatively too many greens from both populations.

We can show that this pr forence has, indeed, little effect on the general frequency-dependent findings described above,. If selection remained constant for the duration of lb and lo then the cross-product rating

no, of browns taken X no. of greens, offered no • of greens taken X no.. of browns offered should be the came in both experiments.. The average C.P.R. gives a measure of the relative preferences: C.P.R. = 0.756 9 which indicates that groans were the favoured prey. Taking this into account we can derive the expected numbers of baits taken. Reference to those values, shown in Table 14, reveals that selection against the rarer morphs is still excessively heavy (x 1 ) 1131.58 0 P 4 0.001 for experiment b and X 2 = 11475.87, P < 0.001 for experiment c).

(ii) Behaviour of the birds

The findings yielded by examination of the grand totals for each experiment are supported by the data collected during the periodic observations. Blackbirds accounted for all the observed predation in is and for most of it in lb and Jo. This is illustrated by Tables 5 and 6 which give the numbers of baits taken per visit during experiments is and Tht lc respectively. Altogether 1 the Observed birds romovd approximately equal proportions from the 1:1 population ends when feeding on the other populations, they tended to take the rare colours, despite an omnipresent preference for greens.

With regard to experiment la, if selection at each visit had been random, we would expect the baits to have been taken in approximately equal proportions. This was clearly not the case.

The results appear to be vary heterogeneous, though they are not - 73 - Tciblo 6

uibezo of baits takon by r10 (o') and Vomalo b1ckbids d bouo oporw=o (11$) dewing obsozvod vioit9 in epoztmonto lb find lo.

ext. lb oxyt , l 9:1 19 Data Tim of Numbem Numbozv obo'vition tabn takon (B.S.T.) G a C B

22.4.68 11.40 5 0 O of 14.43 1 0 23.4.60 16.24 .1 0

20.34 0 7 ? 24.4.38 11.46 2 0 26.4.68 14.45 2 0 O' 30.4.68 10.32 5 0 1.4.68 11.10 2 0 7.4.68 15.25 2 0 0"

15.28 3 0 Q' ' 7V ci IS .39 0 9.4.68 11.00 4 0. ? ci 16.10 2 0 US 10.4168 16.50 0 5 a" 12.4.68 17.30 3 0 cx" 91 2 0 ci 17.42 0 1 US 91 ii 1 OHS ci 1' 0 US cc 1 OH 13.4.60 12.00 1 0 0" 18.10 1 0 cP 14.8.60 10.00 3 0 cx" 15.4.68 11.01 4 0 c, ii 16.52 1 0 0 16 .4 .68 10.03 3 0 10.42 2 0 " 17.4.69 11.15 0 2 0 2 " 11.20 0 3 18.8.68 11.30 8 0 ci if 4 0 'I it 4 0 91 12.13 1 0 US 19.4.68 11.15 0 0 0' 20.4.63 14.43 1 0 US 23.4.68 10.02 9 0 0' " 10.03 3 0 'I 3. OH. ft it 1 OHS 26.8.68 ft 0 ii. O

Czand total 48 17 80 11 a Tho data for 11.30 B.S.?. 18.8.68 ropoort thzoo conoocutive visits by tho ao foto10 blackbird - 74 -

amenable to analysis by chiequered. Indeed, on only ttio of the nine visits did a bird take both types of prey. Thus after selecting

a first colour, a bird would tend to continue picking out the same type.

These differanesa Of behaviour were even more pronounced in the subocquant exporiaents (see Table 6). The results f*om forty..one

visits are shcrn. On average, at each visit 9 we would spact baits to have been taken In the ratio of 48 greens to 17 browns in experiment b and 40 greens to 11 browns in experiment c. In fact not one of the vicits involved predation of both colours. Again we can presume that

once a bird picks out a colour, it continues to select it Even when a bird visited both populations, as on 18.4.68, it still continued to Choose greens from both, although in one population this type was rare and in the other common.

(b) Series 2

The data obtained from these experiments support the main findings described above • Tablo9 7 and B summarise the numbers of baits taken

in experiments a and b respectively and Figs. 7 and 8 present these results graphically by comparing, for each population, the proportions taken tSith the proportions offorod.

The results will firstly be treated by assuming that selection from each population was random., With regard to experiment 2a, it

follows that too many of the rare colours were removed from each of

the four populations containing the mozhs in unequal proportions.. In three of them the effect was highly significant. Taking all four together, there was a significant tendency to take the uncommon colauro (Z •= 23.031, P<0.001). Predation on the 1:1 population - 75 -

Table 7

Cz3nd tota1 of green and bcn etail baits tekon fr'on

pou3.ationc poontod In onporlmnt U. Podato3 tore

Ppno. Pvodatlon Eectod pzdn P I ) 22ed (no 0100tion)

B B

IGIOD 212 362 57.3 516.6 62.66 cO..C31 SGOD 155 3.75 99.0 231.0 45.2S 0.0.001 1021B 260 202 271.0 271.0 0.89 NS 202 99 210.7 90.3 1620 US 901 1B 302 174 500.4 9.6 280.1 <0.001

Thbl 8

totalo of rodand yeUci oa3.1 baits taken from

op1ctiono p000nted in oeint 2b. Pmdators tsoo

b1ackbido and houoe opro,

P £poctod prodm. p Prcpn. Pxodation Ecpoctod pz'e X 1) of2ord (no oo1octio) - (an basis 09

V R V R V

LRs9Y 79 152 23.1 207.9 190.30 40.001 35.1 195.9 6.70 40.001 3Rs77 53.1 61 40 12 93.8 9.82 <0.02 5.8 79.2 0.097 NS R:V 142 08 115.0 115.0 12.60 <0.001 -o _ 7RaSY 84 50 99.8 42.6 7.99 <0.01 112.2 29.8 33.74 '0,001 Dfl1V 260 81 306.9 34.1 71.67 <0.001 319.0 22.0 3.69.40 40.001 - 76 -

was not significantly different from expected.

The birds responded to red and yellow baits in a similar fashion.

Again, in 2b, the four relevant populations all suffered e*coasivs predation of the rare colours and in each came this effect was significant. On the other hand s selection from the isi population

also differed significantly from xpocted.

Apparently, then, selection was f quencydepondmnt in both experiments, that is if we asso that the birds were expected to take the baits in the safle proportions as presented. Ue must now consider the effects of frequency-independent colour-preferences on the significance of those findings. Consider the results obtained

from, the it 1 populations of both experiments. Hero., frequency -

dependent effects should be absent and the values of the cr000produCt

ratios

bG b ) and (,) j( ry should represent the relative preferences for the two colours used in each study. !n 2a the values for b and g are not significantly

different and b 1.076, As can be predicted, when this figure is g used in estimating the expected numbers taken from each population,

the resulting curve scarcely difere from the straight line derived

from assuming random predation (see Fig. 7).

In experiment 2b, however, y a 0.62.

The expected numbers taken, assuming this preference to be constant

from population to population, are given in Table 7 and, as percentages,

In Fig. 8 • Three of the four relevant populations still show over

predation of the colours present at lower frequencies and these

deviations are sal highly significant. The overall tendency to take

Uhevo G. B, R, Y and g, b, r, y represent the numbers of greens, brcmo, rode and yellows offered and taken respectively. 77 -

Figure 7

gil / //

70 C a, ' A a

U) • C 50 // a, .// I- a)

30 /

10 30 50 70 90 /e greens offered

Figure 8

v7 ml 7/

-7- / 7/. 70 a,C /... a /_ 'n50 /./

• 30

ill] 7/ /

10 30 50 70 90 % reds offered

7 and 0 EpoLat 2G1 (cuoll gmow, oo11 bxom8) and 2h (orxU mdo Gu3liL yoUo) MOPOOtIV-03ri parcont0aw Of ono c1 mdo tco iLnot povwntagec offerid. The stajt (bzo!o) tLoo Ltacte the onpactod "Otionchipo mewdua rmd= predation. The cue rapmoon aWetod qmdctlon on tho b&aio o2 the data - 78 -

tho rarcr coloum 5o alec otill eificnt .3o278

/ P 4 0.001).

4. DISCUSSION

The ev.dcnee cbtthod ftora pdatc an 1:1 populationo bea not suøst cy largo differancas in the mlative ettetinoee of ene and bzowue (thcuh rod baits mq7 how) boon profew'od to yeUot). At other fauenoioo the birds, owrall, ton&Dd to take rolativoly

toe cony evocus. Thic izpliee that blockbido, the min pmdatom s favouod emenag a roi3ult that apparently contradiota infoxotLen galnod frorn other eWrimcats (o .g. these doearibod in Chapter VI) carried out with baite at lower denoitioc. B1abiz'e tedod to feed an only cue w1aur durina each vioit

(eeo OzpODiTontQ la, Th and Ic). ThotQb this ouCoetQ that they wera =Ing seGrdbina itoi3ee, tio cannot colotoly rule out the possibility that come or all of the W.rde were coloeting on the basis of proforoncoo otrineio to the oporitxint. Pough (196to l, in sorne e oineuto with Individual chic cdeoo, Pauo Etz'icziilus obtained similar rooulta • T3irde in ooparato cagee tsoro pzeoto4 with a aerico of thirty 1l choicee botoou rod and blvo or yo110 and white ooer seedne The pray wdro offered at danoitie of

100 per 10 inch .actuaro. He found oiificant difforoncee in the behaviour of the hirde • As in the present o oioanto each bird ten&d to pick out seoda of the eate colour es its original choice. - 79 -

Pedctic *a yoJt, cd Mid OMIM=Ov oadt to 1tiooip botoon to pocntoo o f yoliw soods taken by ci cedoor cmd tho porecintaMo oggbmd. Tho otroiGht (b'okon) tho is tho oaVcatod volatianship dctio. WhIlDt tho curm oaooto :axpetod porcontaSeG an tho bavio of date gom 11 opu1atioo. PQU3h 16e).

Ii1

7/ 70 a) ci tn 50 0. -e a) >

10 /

10 30 50 70 90 % yellows offered - 80 -

(b) Selection of ae.p The results Imply that wild birds, on avtrage, tend to select the colour offered in leti froquncy relatively mora often than the cor-men colour. This effect is opposite to that predicted by epoatatic selection, as is well Illustrated by comparing Figs. 7 and 8 with Figs.. 3a and 3b. However the theory of apoetatió. selection originally excluded very high densities-(Clarke 1962a).

For experiments specifically concerned with testing apoatatic selection, the reader is referred to Chapter VI • The maintenance of po1ymorphion by such frequency-dependent selection depends on the assumption that prodators hunting In comploc surroundings, encounter few numbers of prey at the same time. They should consequently acquire searching images that allow for greatest efficiency in exploiting the prey, and those prey-typos that deviate from this image will be protected. Theoretically, these will be the rarest ruorpha. In the present ezperimente the predators could see a large numbar of prey individuals at the same time. Comparable results have been obtained by other workers, notably

Pough (1964). He offered dimorphic populations of painted sunflower seeds to groups of captive chickadees. The size of each population was 100 and the seeds were .randomly distributed on 18-inch square trays. When groups of 6 birds were offered rod and yellow seeds in ratios of 9:1 9. 1:9 and 1:1, they favoured yellows at all frequencies, but also tended to take the rare coloure. Fig. 9 illustrates those results and can be compared directly with Figs. 7 and B. Hayes

(1962 0 quoted by Pough) obtained essentially similar results. Pough's original intention was to test for apootatic ooloction and ho explained his apparently anomalous results by suggesting that the densities had boon too high. - 81 -

The present work with wild birds and Pough'z, with chickadees,

both Support the general theory that when prey era dense predators

tend to osrt 'unifying' (Pie1aaki 1959, 1961) selection. In more

general toresunifying selection, where predators act to o1iainato variability in a prey population, may be regarded as an einp1e of

selection whose effect has been described as 'stabilizing' (Schta1hausen 1949), 'centripetal' (Simpson 1953) and 'normalizing' (Uedd.tngton 1957),

The behavioural cause of such an effect in a prey species is probably quite straightforward. In close groups of prey, any different or odd

types are likely to be conspicuous, whether this difference is due to coloration, location or behaviour. To the human eye, and therefore

presumably to the avian eye (Pumphrey 1961), the rare forms in my experiments appeared to stand out against the background of the rest.

It is reasonable to conclude that the birds tended to take these

types because they were relatively more conspicuous. This is in accordance with out knowledge that pro datora preferentially choose non-cryptic prey (see, e.g., Sumner 1934 and examples quoted by Cott 1940).

Such an overall trend was not the result of homogeneous responses by the birds • On each observed visit in experiments lb and lc,, for example, blackbirds removed either only greens or only browns and, as already mentioned, this behaviour probably depended on which colour was taken first • We can hypothesise that for a given visit this first-eaten bait was most likely to be of the type that was rare and there fore conspicuous. Consequently, on average, this colour was taken relatively more often than the other.

Unifying selection can explain the remarkable uniformity of animals that live in close groups, for example in flocks of birds and - 82 -

shoals of fish • The principle cm also be used to prdict that

polymorphic species should onbibit least variation ( in terms of the number of orphs) at voy higb densities. Polymorphic aniaslc do, indeed, coethiso occur in hiah numbers per =it cujoa ,, At Sanna Bay, Ardasnurchan, it in almost impossible to walk on the sand-dunes without t'oading on Capasa nsizoralis and the density, of haorotnci gcauda in parts of the salt-marshes of the Tyningheme estuary,

East Lothian, may be as great as 10 Isopods per square inch (r.D.J. Heath, personal communication).

Won (1963a, 1965a, 1985b) has studied the e1aticbetween density and the degree of polymorphism in populations of Linicolarje

r'tenaiana. He found that this snail was most polymorphic when it was commonest at dnsitioo of 100 per square metro • In abooluto terms, this figure is comparable to the densities used by Pough.

In effect, hoover, it is probably relatively lower. Consider the two environments. During the experiments, with sunflower seeds and baits, all the prey war's simultaneousiy exposed to the predator's

fIGH of view. They were distributed in a twoisdimensional, unobstructed plane. On the other hand, natural prey mno distributed in many dimensions within complex ouz'z'otxodings. A predator searching for

4mic01aria may, in fact, never oncornter many snails at the same time, oven if they are present in what to W3 scou like high denoitie. Indeed, such conditions may be more suited for apoatatic selection. WO will return (p.2684 to the possible relationships between selection and density, after examining the results of experiments carried out at lower densitio.

S. SUARY

In conclusion, the main results can be ourimarisod as foUotsa: - 83 -

Wild passerines en masse ,tend to select the rarer

varieties of closely packed prey.

This was probably an average result of heter,geneous

responses by individual birds,

The birds appeared to be hunting by searching image.

10 - 8' -

CHAPTER V CODITIOflflG TO GREENS OR BROWNS -85-

CHAPTER V CONDITIONING TO GREENS OR BROWNS

1. INTRODUCTION

The maintenance of colour-polymorphism by apostatic selection implies that predators can become conditioned to searching for one particular type of prey. In Chapter II it was pointed out that that wild predators there is little direct evidence to suggest behave in such a manner. This Chapter deals with attempts at conditioning wild birds to either green or brown baits.

2, PRELIMINARY EXPERIMENT

This early study has been described elsewhere (Allen 1967 9, Allen and Clarke 1968), but is included here for completeness.

(a) Procedures The work was carried out in a garden near Dalkeith, Midlothian

(O,S. grid reference: NT 352678). During six days in September, 1966, approximately 2,000 green baits were scattered on a background of

dark soil marked out in a rectangular grid of forty one-metre squares. A population of forty greens and forty browns was then presented for three consecutive days. Blackbirds were the only predators. After a lapse of one month, the familiarization was repeated

with brown baits, after which the birds were again exposed to a

population containing forty of each colour. During this period dunnooks, house sparrows and robins were feeding at the same time as

blackbirds. It was, therefore, only practicable to observe the latter. Predation by 'small birds' was deduced from examining the

baits at half-hour intervals. Table 9

Preliminary experiment. Daily totals of baits taken after familiarization with one colour alone: soil background.

Hours of Observed predation Date observation Blackbirds "Small birds" (BST) (8+) (many) B GB

Familiarized 6.966 14.30-21.00 50 (9) 9 (1) - -

with greens 7.9.66 11.15-19.00 111(13) 14 (0) -

8.9.66 6.20-18.22 153 (32) 75 - -

Totals 314 (53) 98 - Familiarized with browns 15.10.66 8.25-16.05 52 (8) 103 (20) 110 254

(From Table 3, Allen & Clarke 1968) - 87 -

(b) Results and.diooueaion

The results are given, in Table 9. After each training e3Qoio2 the blackbir& took more of the fii1ier. colour. The trend is highly oiificant despite hstorogonofty bettoen visits • This

point is well made by the figures in paz'ontheeeo in Table 9 9 which record the numbers of visits during which the specified colour uao

taken in etcooe • For each day the figures depart cuLdficently in the expected direction from a 1,1 ratio.

After familiarization with groans there was a significant increase in the percentage of browns taken by blackbirds over every five visits (P < 0.01 9 Spoazcan's rank correlation test), suggesting that the bLrds were gradually rocoieing the browns as acceptable food. During the single day after gaailiarisation with browns there was no corresponding increase in the nuaber of greens taken by blackbirds,, but the 'small birds' did show such a trend (P <0.01).

3. REPEAT EXPERX1ENTS

(a)

Subsequent to the first, three sories of cnálagous exporimento were Implemented. In Outline, they wore planned as followos

Series 1 (tYintor . 1967) comprised four experiments- at separate sites. At two of the locations, birds wore trained on browns and, at the other two, on greens, before being presented with a 1:1 population.

Series 2 (Easter 1969) was identical with series 1, except that it consisted of six experiments.

Series 3 (tYintor 1969) involved two experiments. In one, birds were fad on green baits before being given a 1U population (a). - 88 -

Next, they were trained on browns and again offered a 1 1 population (b). The second experiment was similar, except that browns were od in the first training session (a) and greens in the second (b).

(b) Materials and methoda (I) Locations All the experiments took place on grass fields near Edinburgh. Their sites (nearest farms or villages) and O.S. grid references are given in Table 10.

The skewer-marked eperimenta1 plots were erected near the perimeter of each field, where nearby hcdgoe or tress provided cover for potential predators. Each plot was a square consisting of twenty-five 3t3.-squaroo.

(ii) Procedures At each site, training was completed by scattering it with baits of one Colour, every day for a week. Approximately 2,000 were consumed at each site. On the day following the end of familiarization, populations containing fifty of each colour were presented. These populathne were randomly distributed and were maintained in the usual manner, by repeated replacement (see Chapter III). Table 11 gives the dates of the periods of familiarization and presentation of 1:1 populations. The oxpor'imonts of series 1 and 2 each contained one day's presentation of a 1:1 population, whilst those in series 3 involved offering Iti populations for three days after each training session. tYith one exception, observations were made from a car. from In experiment 12, the birds were watched in the open,! beneath a hedge. - 89 -

Table 10

Sites of opoiznta in oeicc 1, 2 d 3, giving nearest fagwo or villageo cud O.S. grid reftmncoo. All the eoineto wom c&ied out in fielde in Widlothlem. Scetlend.

8R0tN-TRtIWED GREEN-TRAINED Ept • Site Ept. Site Series 1 1.1 Ddon Nino, Roslin. 1.2 3oohiUs, Bwdiohou. TT 27363 NT 268672

1.3 Cpio1ac1 Ro3ooU • 1.1 DThouio Cheetezo, NT 302617 RosoeU. NT 303637

Soioo 2 2.1 Merton. 2.2 Ddon Neine, Rotin. NT 261702 NT 27642

2.3 B'ooiU, Buiehoe. 2.4 Hilleud. NT 260673 NT. 235667

2.5 Upper DcThouoio. 2.8 Thoto. NT 808635 NT 291610

Soiøo 3 3.1e Cpie1Aw, Rosete11. 3Jb (s 3.1a) NT 302617

3.2b (as 3.2&) 3.2a Lihoton. NT 272704 MMM

Tth10 11

Datco of traLyAng and pzooQnQtLon of 1:1 populations in coiinto Qf? oioo 1 2 and S. All dates am ino1Lv.

BROMmTRAINED .SRAXI1ED

Ept, Dcteü of Dtoo of Eapt. Diteo of Doü of training p36ontQt10 tiItnj3 pontatio of 110. of 11jo.

Sewjol3 1 1.1 12.11.67'. 19.11.67 1.2 18.11.67.. 20.11.67 18.11.67 19.11.67

1.3 90.11.67. 7.12,67 1.4 8.12.67' 16.124.67 6.12.67 15.12.67

So1oo 2 2.1 4.3.63'. 11.3.68 2.2 4.3.68'. 12.3.68 10.3.66 10.3.68

2.3 9.3.68.. 15.3.68 2.4 19.3.63'. 26.3.68

• 15.3.63 25..3.G8

2.5 18.3.68'. 25.3.68 2.6 24.3.68'. 31.3.63 24.3.68 30.3.66

Sioo 3 $.lc 20.11.69.. 28.11.69'. 3.1b 31.11.69- 7.12.69..

• 27.11.69 30.11.69 6.12.69 8.12.69

3.2b 13.12.69.. 19.12.69'. 3.1.0 3.12.69'. 10.12.69'.

18.12.69 21.12.69 • 9.12.69 12.12.69 - 91 -

Table 12 'S Humbew of baits taken from 111 populations by blackbirds

and house sparrows () after training, and estimated numbers

of birds involved (8hoin in parentheses).

BROWN-TRAINED GREEN-TRAINED

ozcpt. Nuere taken caxpt. Numbers taken B C B

Series 1 1.1 1 65 (2) 1.2 26 0 (2)

113 0 14 1.4 55 0 (2)

S3ieQ 2 2.1 0 111 (4) 2.2 92 0 (2)

2 2.3 0 51. (8) 2.4 86 (2) 0 31 (6)

2.5 0 58 2.6 41 0 (2)

Series 3 3.la 0 71 (2) 3.1b 129 81 (2)

3.2b 24 135 (8) 3.2a 106 0 (4)

Pwilminary 52 103 314 98 eWriment (80 (8+) - 92 -

(iii) Predators Apart from one experiment, blackbirds wra the only predators noted feeding. In experiment 2.3 house sparrows were also involved.

The'h numbers of baits taken by these birds were estimated as in, the preliminary eperimant. In each of the present studia, compared with the preliminary experiment fewer individual blackbirds were involved. This was probably a result of the differences in enironments; blackbirds are more abundant in gardens and woods than in open fields and hedgerows (Snow 1958). Estimates of the numbers of blackbirds participating in ech experiment are given in Table 12.

(c) Results Table 12 gives the numbers of greens and browns taken during the presentation, of 1:1 populations. The data for each experiment in the first two series represent the results from a single day's prosentation, whilst those of series 3 are grand totals of predation over three days. In addition to the results of eorios 1, 2 and

30 Table 12 gives the data from the preliminary experiment for easy refer-ace. In oil, fourteen populations were presented to birds after

separate training regimes. The effects of training are clear-cut- In each population s, the familiar colour has been predated to excess. That is to oay, during presentation of 1:1 populations, birds wore continuing to search for, and eat, those baits that wore the sacs colour as the ones on which they had been trained. - 93 -

Tthio 13

pOZDitO 3.1 md 3.2 . Doily po&i@

by b1bidO.

poIct 3.10 Epii.tot 3. lb

BRO1N-TAflED GREEtTRAXNED

Date Nunbors taken Data Humbera to (1969) G B (1969) G B

28 NOV O 24 7 Dc 53 9

29ov 0 33 8Dec 39 114

30 Nov 0 12 9 W 37 16

0 71 129 131

Epimmt 3.2b EzpezitQnt 3.20

BRO..TRAINED GREEN-TRAINED

Date Humberc taken Data Numbers taken (1969) G B (1969) G8

10 Dc 5 56 19 Dec 35 0

11 Doe 6 - 36 20 Dec 44 0

12 Dec 13 43 21 DeC 27 0

24 135 lOG 0 * 94 -

A further measure of the acquired preferences is given by the

fact that in ten populations the unfamiliar colour was totally neglected, whilst in the remainder only one and two were taken

(poziments 1.1 and 2.e respectively). What is more, the unfamiliar

colour was totally untouched in each of the three days of 1:1 presentation in experiments 3.1a and 3.2a, after the first training sosolons. In other words, in twalvo experiments there was a homogeneous response to the prey • It is legitimat% therefore, to compare otetioticaliy the observed and expected values of the proportions taken. The deviations from expected 1:1 selection are all highly statistically significant. Each has a probability of less than 0.001 that the result was due to chance alone.

The only experiments that revealed appreciable elimination of the uncestcmer'y varieties were 3.1b and 3.2b. Presentation of 1*1 populations after training in these studies had already been preceded by experiments 3.1a and 3.2a respectively;, The birds were therefore familiar with both colours by the time the second training period was implemented. Vet by the time of the subsequent presentation of 1:1 populations, they were apparently again conditioned to searching for one particular colour, but now the varieties involved were the ones that had been

'unfamiliar' in the first experiment. They continued to take an excess of these colours on each of the three days of presentation in each experiment (Table 13) • Nevertheless, in one experiment there was a significant incrociso in the proportions of the unfamiliar colour taken over ovejr three visits, suggesting that the birds were ro..leerning to search for these prey (r5 0.739 P 40.05 w for experimefl 3.11 0.69, not significant., for experiment 3.2b). In addition, there were significant differences in the proportions taken on the first and last days after the second training sessions, but not between any other

1:f - 95 - combinations of days 5,44147, P < 0.02 for experiment 3.1b; Xl) 5.057 P < 0.05 for experiment 3.2bL

The results for the two experiments of series 3 imply that preferences produced by 'a first session of familiarization can be changed by a second. Though it was difficult to identify reliably

unringed individuals, there was no evidence for changes in the compositions of the predators after any of the training periods. It is therefore reasonable to assume that individual birds underwent changes of preference. This argument is strengthened by observations

on a blackbird with two conspicuous white feathers in her tail. The numbers of baits taken by her in experiment 3.1 are shown in

Table 14.

Table 114

Numbers of baits taken from 1:1 population by a recognisable female blackbird in experiment 3.1

Numbers 'eaken G B

After brown-training 28th Nov. - 30th Nov. inclusive 0 24

After green-training 7th Dec. - 9th Dec. inclusive 43 - 96

;. DISCUSSIOI'1

('a) Comparison bat-Veen preliminaz and ur.eh6 att % eiients

All the above experiments apparently demonstrate the striking

effects of training. It will be noted, however,, that the results of the duplicate experiments are more explicit than those of the preliminary study. This small discrepancy may be a function of the

differences between the two typos of environments: fields and

hedgerows as opposed to gardens and woods • The 'former habitats,. support relatively fewer blackbLrds (Snow 1958) and the numbers involved

in the duplicate experiments were correspondingly lower. Probably as a consequence of their low densities, the compositions of blackbirds appeared to remain stable 'and confined to those with territories 1 in the immediate vicinities of the experiments. Thus each bird had a

high chance of feeding on baits offered during training.

Gardens and woods, in contrast, support a higher density of blackbirds, and throughout the pilot experiment it seemed subjectively

that new birds were periodically finding, the experimental food supply. Conaequently,some of the birds observed feeding on the l:lpulationa

may have eaten few, or none, of the baits offered during training. This could explain why some birds took unfamiliar baits, sometimes

oven to excess.

(b) Effect of natural Breferencoo .

So far it has been assumed that green and brown baits are normally equally acceptable to birds • Hence the expected selection

from the dimorphic populations was assumed to be IU • In fact,

1Tho relevance of territory and territorial behaviour are diacused in Chapter VIII, in relation to some 9:1 experiments.. - 97 - blackbirds Conerally soom to pz'sfog borno (ese Choptss VI VIII) end this natural predilection oo8d conoivsb1y account for the roaulta obtained after fmiiciition with this colour. Hwevor g zneo to enporimnto described in Chapter T uides this possibility tmlikely. Tho proscatation of dimorphic pop1ations •Lth bne nine tir.ee as eoon sa ens Savo rosulto for naive blackbirds (end OZio congthh) on six separate occasions (see Table 20) • The grand totals taken by these birds tevo 13 greens and 2002 bras a ratio of approulmataly WOO. In the brotmatrsintn oiaouts described above (neglecting the preliminary oXpori!snt

Md 3. 2be both of which invold an extra proiouo troining 1 grcn and 370 brot'ns ware taken, That is to say., the pportioue of brosno and greens taken we= almost twice as groat in the present 1*1 aWrimcrnta as in the specified 9l etwUes. Yet in the lattor, browns ware presented in ralativaly greater noro • It is therefore z'oasonthlo to conclude that brown training does in fact reinforce any natural predilections for this colour,

(0) Condition

Szarising the rMin results, wild birds continued to search or familiar colours after each of sixteen separate training sessions. These conclusions will now be discussed with regard to the findings of other workers and, briefly, in the content of results reported esowhore in this thesis.

The repeatability of thoo oxoricnta is lent further waight by sots similar work, using green and brown baits, done at Liverpool. by Sister S. Shutt (personal eoaunicstion) • She fed blackbirds (no other birds wore inlvsd) on green lard.iendflour pellets for a week, on a backgrotmd of rod tiles. Twenty browns and twenty groans wore - 98 - then presented at an unspecified density. Baits were not replaced when eaten and only a pair of birds were involved. On the first day, seventeen greens and only one brOtm were taken and on the next, only greens. On each of the succeeding two days the male blackbird was the first to visit the populations and took greens alone • Confronted with a scarcity of her preferred colour, the female was forced to commence eating browns. Similar results for wild birds have been obtained by Oates and others (quoted by Cook 1971, pp.87a88). The prey were red and yellow 'maggots! made of flour, fat, and colouring and they were presented in populations of 49, at a density of one per square foot. Predators were bousesparrows, blackbirds and starlings. After four days of training on one colour, both types were presented for four more days. Their proportions were either 90149 70% or SOi of the conditioning colour. In ten out of twelve trials, a relative excess of the conditioned colour was taken (see Cook 1971 9 Table 4.1). As in my data, there was a suggestion that the tendency to remove the familiar colour declines with time. It seems probable that the effect described in this Chapter is widespread in wild birds. Including myself, three authors (and probably others) have obtained similar results • Our data support

Tinbergon 's (1960) original contention that wild birds can build up searching images to prey and may consequently overlook new species,, oven if those are relatively common. Or',tf the two prey were morphs of a species, then we have a mechanism capable of taintaining the polymorphism. Further supporting evidence comes from experiments with captive birds. The work of Rabinowitch (1968) has already been described (p.3), but considered unsatisfectorg since ho used chicks of a very early age. Pough (1964), however, was able to demonstrate •0

the effects of conditioning on prey selection by captive chickadees. Birds were fed on either 100 yellow or 100 red sunflower seeds for two days before being presented with 70 of the familiar colour and

30 of the other. In other words, he offered 9:1 populationa overall as in his other experiments (see Chapter IV, pp. 78-80) except that the birds were obliged to encounter 100 of the common sort before presentation of the remainder of the population. In comparison with his earlier experiments, he found that selection was now closer to expected, assuming random predation. Presumably, the tendency to take rarer forms was being balanced by a conditioned preference for the commoner type. More recently, similar results have been obtained by Mrs. Pat Miller (personal communication) for quails feeding on red and blue pastry prey.

Other results comparable to mine have been found for fish

(Reighard 1908, Beukema 1968, Smith 1967) and mammals (Soane 1970).

Most of them (and also those with birds involved morphs that are visibly, very distinct. Green and brown baits are no exceptions to this rule. The morphs of many species are often similar in colour. Croze

(1967, 1970) is the only worker to have used visually similar artificial morphs and his findings will be discussed later, in conjunction with experiments of my own. The evidence presented in that section (Chapter IX)will lend more weight to the present findings. More support is derived from some of the findings described in the next Chapter, which deals with the responses of untrained birds when confronted with populations containing greens and browns in the proportions of nine to one or one to nine. - 100 -

S. SU'1ARY

In ccnc1uion, we can summarise the main results as follows:

In 8 experiments, idld birds tors trained to search for groan or bro'm baits.

t1hGn presented with a mixture of both In equal proportions they took the familiar colour to excess • This occurred in all experiments.

In 3 experiments it was possible to reverse the preference by carrying out a second period of training.

The results of the same 3 experiments indicated that the effect of conditioning gradually 4esoened with time. - 10]. -

CHAPTER VI 9:1 EXPERIMENTS: PREDATION BY GROUPS OF BIRDS - I o—

CHAPTER VI 9:1 EXPERIMENTS: PREDATION BY GROUPS OF BIRDS

1, INTRODUCTION

This section deals with attempts to determine whether 02H2sl of naive birds will proforontially search foz the Coummonor colours

In populations of groom and brown baits • The background to this line of enquiry has been discussed in Chapter II (pagss41 47). At first it was decided to present the baits in populations with one colour nine times as common as the other and later oporinments adhered to these proportions. Each experiment consisted of two parts. Birds wore first presented with populations having one of the colours nine times as cowmen as the other. After a few days they wore offered control populations with the other colour as the common variety. Such a design meat that two basic types of experiments wore carried outs they started with populations of either 9 greens: 1 brown, or 1 green: 9 browns.

2. PRELIMINARY EXPERIMENTS

Two of these studies were carried out as part of a B.Sc.

Honours project (Allan 1967). The results of the first experiment have since been further analysed and have been published (Allen and Clarke 1968).

(a) Exi,oriment 1: iresontation of 9G:1B poDuations followed by 1G9B popu]ations (I) Materials and methods

A lOm X lOm grid was laid on the lam outside the Department of

1This does not necessarily imply that the birds wore feeding in the same place at the same time. - 103 -

1c j9. 1. Da-Uy total of belttalcons

Table 1. DAILY TOTALS OF BAITS TAKEN: GRASS BACKGROUND Experiment 1 180 greens : 20 browns Hours of Observed predation Date observation Blackbirds Starlings Dun- H. Sparrows (BST) (6+) (2) flocks (2) (many) G B G B GB G B

30.7.66 9.03-1.0O 2 0 - - - - 1.8.66 10.30-18.15 61 9 0 1 1 0 - - 2.8.66 8.30-16.30 71 22 - 4 0 - - 3.8.66 9.20-15.15 71 17 - - 5 0 10 1 5.8.66 10.30-17.05 45 11 - - 14 0 86 2 6.8.86 5.09- 7.51 69 15 - - - 14 0 7.8.66 16.06-18.10 35 12 20 2 - - 57 0 Grand totals 354 86 20 3 24 0 167 3 Expected 9 : 1 396•0 440 207 23 216 24 153•0 170 Expected, constant selection 3490 910 152 78 234 06 154•4 15•6 Experiment 2 20 greens: 180 browns Hours of Observed predation observation Blackbirds Starlings Dun- H. sparrows (BST) (2+) (many) flocks (2) (many) Date 0 B 0 BOB 0 B- ___w 9.8.66 9.30-15.15 . - 0 7 0 1 15 42

10.8.66 10.30-17.30 0 5 - - - - 5 155

11.8.66 10.50-16.10 0 14 1 37 - - 14 183 14.8.66 15.05-18.10 0 13 0 45 0 2 4 136

15.8.66 11.40-14.10 0 13 0 44 - - 2 80

17.8.66 14.30-18.35 1 11 1 19 - - 9 74 18.8.66 10.25-17.40 0 24 1 36 .1 4 29 111 Grand totals 1 80 3 188 1 7 78 761 Expected 1 : 9 81 729 191 1689 08 72 839 7551 an.. 1jpvuvvu, .uI1 - stantselection* 3.7 77.3 45 1865 28 52 914 7476 * See text. The approximate numbers of birds involved In each experiment are given I at the heads of the columns.

Mood on Tb10 10 E11on and C1rco 198) - 1014 -

Zoology. Univorsity of Edinburgh. 1ithin the grid, 180 green and 20 brlctin baits wore distributed at random, two baits per metre square.

Observations were made from a first-floor windw and the population was maintained in the usual manner. During the experiment the grid es under continuous observation and a record was kept of tho numbers of baits token by the birds at each visit. After seven days of predation, the otperiment was repeated with a population of 20 greens and 180 brct4ns.

(Li) Results and jCcuacion The daily totals of baits eaten are shown In Table 15. Below the grand totals are given the expected numbers based on the assumption that there is no selection. The blackbirds and starlings took more browns than expected in both experiments. Dunnocks and sparrows, on the other hand, took more greens In the first experiment and more browns in the second. It is c1ear, therefore, that the birds did exercise visual selection and the direction of selection

could change within a few days. These changes did not seem to be caused by alterations in the colour of the background, which remained

apparently constant throughout the period of study.

It romaine to enquire whether the differences between the two experiments are related to the frequencies of the colours. On the basis of thGirand totals we can calculate for each species the expected numbers of prey eaton assuming that the selection remained constant throughout the period of study. If this were so the cross-

product. ratio:

ITo • of areons offered X 1o. of browns eaten Lo. of brownsofforod X No. of greens eaten should be the same in both parts of the experiment. - 105 -

TcMqj. oitnt 1. Numbora of bito tdcon by bla&-birft ding Ldivdu1 vio em 1uut 2 9 1966.

No. of baits taken Visit U. B 1 4 2 7 0 3 3• 0 4 3 0 5 9 0 6 5 0 7 5 0 8 2 0 9 o •io 10 24 0 ii 3 0 12 5 0 13 1 0 14 0 12 Total 71 22

(Pmm Tablo 2e Allan and Clca .968) - 106 -

The expected numbers based on assuming constant selection in both experiments, are given in the second row below the grand totals.

It can be seen that, in fact, all four species took a larger number of coznon forms than expected and a smaller number of rare ones.

Thus the birds appeared to select in a frequency-dependent manner.

Direct comparison of observed and expected numbers suggests that the effect is highly significant (()2 cs 20.19, P < 0.0001)) - This

comparison, however, is not strictly legitimate because of heterogeneity within the series, deriving from two aources Heteenoitr between days. For example, in the oocond part,

house sparrows took the two colours in different proportions on

different days(6) 59.15 0 P < 0.001). HeterogeneitY between visits • Even where there is no significant

heterogeneity between days, as with the rsulto for blackbirds, there

is strong evidence that the proportions differed from visit to visit. Table 16 shows the numbers of green and brown baits taken by blackbirds

on August 2nd, 1966 • Each visit by an individual bird is recorded

separately. The data are not amenable to analysis by chi-aquaz'ed, but it is clear that they are very significantly heterogeneous. If

the birds wore taking the baits at random, but encountering then in the proportions of seventy-one greens to twenty-two browns, than the

probability of a bird taking twelve browns in succession (as on visit

14) would be loss than 1 in 10. When this heterogeneity is taken into accounts the frequency"

dependent effects, although striking, cannot be regarded as formally

significant. - 107 -

Table 17 ELt 2.1. DaIU totqI6 of bafto ton

Pdatov two blackbirds one swgtbrmh

a. 20 gonoz 100 bron

Data Hour@ of obsavation 0boeod po&itio (1967) (G.1.T.)

18 1 1 7.48 16.30 1 58 19.1 7.50 16.51 o 49 20.1 7.40 a 17.00 o 38 21.1 8.25 17.00 0 54 22.1 7.35 17.00 0 38 23.1 8.20 17.00 o 55. G'aud totals 1 (1.1) (311.9) b • 180 gmaws 20 bno

Data Hours of observation Observed pdrthn (1967) (G.4.T.) C B

24.1 8.35 • 17.00 6 39 251 8.10 - 17.00 18 40 26.1 8.07 - 17.00 17 37 27.1 8.14 17.00 5 33 28.1 8.3,5-17.00 9 45 29.1 9.00 - 17.00 12 36 30 11 8.30 17.00 16 50 31.1 7.50 - 17.15 9 47 Grand totalo 92 327 (91.5) (327.5)

(Numbers in parentheses represent the grand totals expected on the assumption of constant selection). - 108 -

(b) Experiment 2.11: presentation of 1G-.9B populations followed by 99:1B populations

(I) t4aterials3 and methods

The predators involved in both parts Of this experiment were

two blackbirds and a thrush. All wore colour-ringed and results were obtained for predation by each one. Hero we shall consider

their combined predation, whilst their individual responses will be examined in Chapter VII.

The procedures are described fully in the next Chapter. umtharising.'thèm, on 16th January 1967 a 10n X lOIi grid was. laid down on a lawn at my Reading home • For six days the birds were presented with populations Containing 20 greens and 180 browns. They were then offered 180 greens and 20 browns for a further 8 days. Other populations were then presented but these do not concern us hero

(see p.151) • All the populations were continuously observed and they were all presented during most of the daylight hours (ace Table 17).

(ii) Results and discussion

Table 17 gives the pooled daily results for predation by the three Turdidao • Too many browns were clearly taken from the two types of population. If this preference is csswaed to be frequency- Independent and constant in both parts of the study we can derive the expected values shown in parentheses. The magnitudesof the deviations are clearly negligible.

The birds searched mainly for browns after the relative Composition of the baits was changed. These d.y results were not significantly

1Exporimezits 2.2 and 2.3 are described in Chapter VIII. Table 18 EQirnot 13, 22LU , totaU of ittkon

Date Hoa of Blackbirds Docko Robins • Bltm Tits } observation (2-4) (2) (2) (2) 0 BG. B 0 B C fl C , 20 oo: 2.12 12.15 - 16,20 0 21 - 1 2 - 9 180 bmune 3.12 9.15 - 16.30 0 26 - •- 1 4 - 23 4.12 9.15 - 16.45 o 35 0 9 - •- - - 16 80 5.12 9.25 - 16.80 0 29 0 2. 0 4 0 8 13 01 6.12 9.35 16.30 0 67 1 15 0 5 0 7 3 59 7.12 9.41 - 16.20 59 0 13 0 6 0 11 11 53 Girnd totaló o 232.1 39 2 21 0 26 47 296 Epect 1:9 23.2 208.8 3.5 2.3 20.7 2.6 23.4 33 300.7 Epectod, ci tct oo1otion 3.3 228.7 1.4 33.6 2.2 20.1 0.3 25.7 50,3 293.7 b, 100 amoum 9.12 9.35 - 16.25 16 50 1 2 3 1 3. 5 3.0 30 - 47 0 3 1 2 25 20 103 broio 10.12 9'.90 3.6.50 U 4 5 0 99 7 9 0 2 1 2 16 U (0 11.12 9.20 - 16.50 12 5 12.12 9.15 - 16.40 7 48 7 9 3. 3 2 0 29 Grand tot10 46 19*4 19 25 9 U 5 9 80 122 Zpeted, 1:1 120.0 120.0 22.0 22.0 7.5 7.5 7.0 7.0 101.0 101.0 Ezoctod. c.&t3nt selection 27.7 212.4 12.0 32.0 7.2 7.8 1.1 12.9 121.4 8016 e, 100 SmOaGs 18.12 9.40 -• 16.55 32 20 15 5 4 0 3 1 68 20 bzc 14.12 9.20 - 15.30 44 33 14 6 2 1 - - 90 3 OL 15.12 9.37 - 16.10 91 21 12 4 0 - •- 37 0 Grand totals 117 74 41 11 10 1 3 1 195 9 Epctod, 9:1 171.9 19.1 4648 5.2 9.9,.. 0.1 3.6 0.9 130.6 15.4 Epocto, ecnott aelectim 103.1 87.9 90.2 11.8 9.8 ^ 1.2 1.7 2.3 193,4 10.6 - 110 - heterogeneous (X 5) .3O, P > 0.50) despite variation of the individual responses. Tuo birds searched for browns almost exclusively when rare, but the other relaxed its preference On the first day after changeover. We shall return to these birds in the next Chapter.

This expoimant, carried out in the late winter of 1968 9 was essentially a repeat of experiment 2.1, except that 1:1 populations were offered between the ii 9 and 9:1 populations e-

(1) materials and methods

On 2nd December 1968 a grid of 100 metre quarea was laid down on the lawn of a garden in Dalkeith. For seven days birds were presented with 20 greens and 180 browns. After a lapse Of one day, equal numbers of greens and browns were offered for four days., followed by three days of 180 greens and 20 browns.

The baits when presented were observed continuously; blackbirds and house sparrows were the main predators but robins, dwmnocks and blue tits were also involved in all three parts of the study. In addition, later populations. were visited by great tits (Parue major I.), coal tit (P. atop L.) and starlings.

Results and discussion Table 18 shows the daily totals of baits taken The only results that reveal significant heterogeneity in the proportions taken per day are those for house sparrows in pert b. 3) 16.93,

P <0.001). As in experiment 1, however, the blackbirds exhibited considerable variation in the proportions taken per visit when feeding on populations presented during b a and c. - ill -

Below the grand totals are given the expected numbers based on random predation. The birds were clearly discriminating between the baits; thus the blackbirds, dmnocks, robins and blue tits took more brans than expected in all three parts. The house sparrows took too many greens in a. and c. and too many browns in b. The Table also gives the expected predation baod on the assution that selection was constant in a • and c • Those values wore dst,od in the same manner as before C see page 1044. In all ton cases the commoner colour was taken more often than expected. Overall this affect is statistically significant 5.7 9 P C 0 102). (( n The data from the 1:1 populations need some covmont • In every case browns were taken to excess on the basis of ll ratios, suggesting that the birds carriod'-ovot' preferences loareod from part (a) of the experiment. Overall this effect is statistically highly significant ((W2 53.07 9 P t o.00i) . On the basis of constant selection in parts (a) and (b) we find that robins and house sparrows took more browns than expected (X 1) 2.8030 P ) 0.05 0 not significant; X2 33.399 P < 0.001) • On the Other hand the 1) blackbirds and dunnocke for some reason took more greens than oXp6ated (X 13.76, p

3. REPEAT EXPERIMENTS

,Two of the three continuously observed experiments apparently showed statistically significant frequonc3r.'.dopondont solecticn. However, all the studies revealed ccnsidorble heterogeneity in their results and it must be re.-ephasisod that the use of chi-oquarad was - 112 -

therefore not strictly legitimate. This problem could be overcome,

to some extent, by seeing whether repeat eporiments give similar

evidence for spostatic selection. Consequently, eleven more

experiments were carried out: six followed the design of experiment

1 and five were similar to experiment 2. During these experiments

the birds were not always observed.

(a) Plan of %ork

The experiments were grouped into four separate series carried out in 1968 and 1969. These series were designed as. follows.

U) Series 3 August 1969 Four experiments were involved; two were carried out on grass and two on soil. All were at sites within the City of

Edinburgh and were lees than one mile from the Department of Zoology. The population were not always observed continuously: occasionally

they were left unattended and ro..oxainiuod on return • Birds could therefore be fed at more than one site on the same day. Each period of predation, observed or unobserved, constituted a 'trial'. The populations contained 100 baits laid out in 5m x Sm grids. Ideally they should have been visited as frequently as osciblo and

it was hoped to keep the recorded predation to bow 50%. This was not always possible, because of a combination of high rates of predation and transport diffcultiea. Larger populations may have helped but were ccoidered impractical (but coo series 5 and 6).

(ii) Series 5 February to April 1969

The six experiments of this series were similar in basic design to thococr series 3. However, all were carried out on fields

In Ridlothian and they involved populations of 200 baits distributed

In lOs x lOs grids. Those larger populations and the use of a car for - 113.-•

getting from site to site ensured that predation could be kept to

within reasonable limits.

The start of the series was delayed by heavy falls of snow during

the first to weeks of February. During the middle of the month,

the sites were almost clear of snow and the first experiments started

On the 14th • On the same day more snow fell and the plots remained

completely covered until the and of the month. A gradual thaw sot in

at the beginning of March and all the sites were completely clear by the 5th.

(iii) Series 6 - Nay to June 1969

This set contained 2 experiments carried out in grass fields.

The populations were distributed in iOn x iOn grids and, like those Of

the experiments in the other series, wore not under continuous observation.

Rain fell, on various occasions and was particularly heavy on 6th and 14th May and on 21st and 22nd Juno.

(b) EEperiments and results

These experiments were studies unto themselves and, like the first three, demand individual attention. However it is felt that the insertion of these bulky results in the ensuing pages would interrupt the present theme • The reader should therefore refer to Appendix A. pages i3 to 146 for full details of the eleven repeat experiments.

4. SUMMARY OF DATA

Tables 19 and 20 summarise the data. They give the grand totals of baits taken from the two typos of populations presented in all the experiments described in this chapter. Data for predation on 90:1B * 114 -

Table 19 -

Gmad t0t013 of balts takca In 981 a oLt2ont 9( 113 pop u2.Qtiono ioilosod by 3G 99 populations. P&to b1ackbidü (1) ootnoh (So),, house spoEwue (HS) o duo& (D), ct 1.ng (St)0

8 19 98

Ept. Pzd. G B 0 B CIPR X 1)

1 HS 167 3 78 761 : 0.907 (154.4) (15.6) (93.4) (747.6)

D 24 0 1 7 0.21 20.19 40.001 (23.8) (0.6) (2.8) (52) (ux) 2')

St 20 3 3 188 4.89 (15.2) (7.8) (8.5) (186.5)

91 358 86 1 80 2,81 (348.9) (91.1) (3.7) (77.3)

3.1 BI OSO O 363 120 1 202 3.46 6.63 40.02 HS (349.1) (133.9) (663) (196.7)

3.8 B1St 9 128 47 1 262 5.00 9.00 40.01 HS (112.5) (62.5) (5.7) (257.3)

5.2 21 353 56 31 278 1.73 4.32 40.05 (343.0) (66.0) (17.4) (271.6)

5.3 21 231 40 7 299 2.24 8.60 40.01 (217.0) (54.0) (18.5) (291.5)

5.4 Dl 192 39 6 388 2.84 10.82 40.01 (175.6) (55.4) (13.6) (340.4)

6.2 81 619 188 5 333 2,48 8.13 40.01 (603.3) (165.7) (13.6) (304.4) - 115 - Table 20

Grand tQtEal3 of baito taken in 933. atprimnts with 1C9B populationa followed by 9G-,),B popiation0 Pdator: b1ackbizd8 (BI) q songthruoh (So) hoo sparrows (HS), &tnnek& M. robina M.

9B >9C : 13

Ept. Pxnd. GB G B CPR X 1 ) P 21 B1,So 1 312 92 327 32.23 0.01 not sign. (1.3.) (311.9) (91.5) (327.5)

MS 47 296 145 9 0.664 (51.3) (293.7) (133.4) (10.6)

D 1 34 41 11 2.66 (1.4) (33.6) (40.2) (11.8) 574 4042 '-' 2) R 2 21 10 1 1.07 (2.2) (20.1) (9.8) (1.2)

BI 0 232 117 74 7.68 (3.3) (228.7) (103.1) (87.9)

3.2 BI 2 734 63 102 20.40 5.44 40.02 (.0•) (732.0) (50.5) (114.5)

3.3 31St. - 2 287 167 139 8.31- 3.35 not oi

MS (5.8) - (283.2) (3.59.1) (146.9)

5.1 BI 3 368 372 90 2.95 15.18 -0.003. (13.5) (357.5) (348.0) (11.0)

5.5 31 3 350 380 Si. 2.65 16.25 <0.001 (14.2) (338.8) (356.1) (104.9)

611 31 4 606 411 131 3.99 20.72 <0.001 (16.5) (593.5) (375.4) (166.6) -fl6-

populations followed by 1G: 9B populations are shown in Table 19 9 whilst Table 20 presents the results from the complementary types of experiments tthore possible (i.e. in expts • 1 and k) predation by different species is shown separately. This means that the Tables show 20 sets of data from 14 experiments 1 .

Cross-product ratios were derived by assuming that selection was constant in both parts of the experiments and are shown in the Tables.

The numbers In parentheses represent the expected predation and were calculated on the basis of the cross-product ratios. Also shown are the values of chi-squared for the deviations of observed from expected predation and the probabilities of these deviations being due to chance alone.

5. DISCUSSION

The discussion naturally centres on the evidence for apostatic selection. We also examine the frequency-independont colour preferences and discuss the direct evidence suggesting that the birds can become conditioned to searching for familiar colours.

(a) Apostatic selection Reference to Tables 19 and 20 will reveal that most of the grand totals deviate from expected in the directions predicted by apostatic selection. Of the 20 sets of data, 19 exhibit over-predation of the common colours and under-predation of the rare ones, in both parts of the experiments. This proportion of scores is highly significant

(P< 1idmia1 test) and overcomes the problem of heterogeneity within sets.

1The data for blackbirds have now been published (Allen 1972). - 117 -

In terns of onporimnts,, 12 out of 14 revealed statistically 3ififioant deviations from expected predation aaou4ag constant selection. As can be seen from Table 19, there was apparently eiifLcant frequency -dependent selection in all 7 exporimenta of

the type 9G: 13 • 3J3: 93. The findings of experiment 1 therefore seem to be repeatabLe. when populations were presented in the revso sequence (see Table 20) 9 the results of S of the 7 porinte ohThitod siguifleant apoetatic seloøtion • The duplicate eperieente did not reveal the existence of such strong brown preferences as wore apparent in experiment 2.1, but the tendency was 3oTetivnoe present (see below). The eperirnonts encompassed a variety of eiporiental conditions.

Some took place in winter; On short grass (2.1) 0 or against a background of snow (parts of 5.1 2, 5.2 and 5.3) • Unftr the letter conditions tie night expect selection to be alone to random (see ,118 , Other studies were carried out in spring (5.4, 5.5), and auner (all of series 3 and 6). The experimental areas had various

backgromds; soil (3.2 9 3.4), short uniform grass (1, 2.1, 3.1 0 3.2 0 4 9 5.5 1, 6.2) or longer patchy, grass (5.1, 5.2 9 5.39 5.4.0 6.1). Different birds were involved at different sites. There was also heterogeneity within experiments. Besides preliminary experiment 1, only experiments 5.2 and 3.4 did not show significant heterogeneity of the proportions of greens and browns taken per trial when birds were feeding on 9CslB populations. Yet the overall results are remarkably alike, despite the dissimilarities of the experimental conditions • There was always (with the exception of 2.4) a tendency for the 0omcn colour to be taken core often than expected on the basis of constant selection. The exporimento therefore appear to be consistently repeatable. - 118 -

(b) Colour preencee The preliminary experiments took place on lawns • In expez!imnt

1 the Turdidae tended to take too many browns from both typos of population, assuming 9:1 expectations, and this observation was supported to avon gOQtOV affect by the second experiment. Subsequent experiments (3.1 9 3.2 9 49 5,13 9 5.5 9 6.1 9 6.2) on grass backgrounds also gave similar results for blackbirds and thrushes. During the presentation of 10 greens and 90 browns in experiment 3.2, for example, the browns wara completely eliminated during 4 of the 10 trials, whilst apart for the losE, of one bait, the greens were untouched. The average cross.-product ratios derived from the grand totals of the

results of those experiments are shown In the relevant Tables: all

are greater than unity. An explanation for the preferences of the Turdidas could be tht

the browns wore less cryptic than greens against grass. To the han eye, however, this did not appear to be the case. 'What is more, the results of other experiments indicate that brown preferences are

independent of the colour of the background. Snow covered the ground during many of the trials of experiments

5.10 5.2 and 5.3. In comparison with graze, this surface was more uniform with regard to both colour and composition. Against it greens

and browns appeared to be wquaLly highly conspicuous. If the colour pro forenoon were a function of the background coloration we might expect relatively fwr browns to have been taken in those oxperimonts

than in these on grass • Another factor might also be expected to contrlbuto to making selection random. Various workers (Young 198,

Prop 1960 Boukama 1968) have shown that predators appear to relax

their fcod..preferonctas when hungry, as when their normal environments 119

zra eovod by anow. Rofomneo to tho moulto of the tbx 09c3%t

(pp. 110 141 pp. 128 129 0 qp.130 131) ohzoo that the ?1.&o ot11 to coo bma thca ato on baolz'og tho poto

Ppt11QtLc!0 pr000nted an soil (OUPOVAMOSto 3.3 c.d 3•t) pzduco1 oou1to comparabla. to thoo luot diacuasod o but for the =blnod prodatlw by a vdatm of dLffoont apooLor. Thourh ameno c9powod

mom acnoplatnus,too ty brwoo too tikcii fz both typoa of

Mulatien than opootd by octic. Iblo pz'ofoao was no poo to dayn in 3.3 that aU the cam= bona tov

taken thiot the gm ow teo loft ad (ace ?ab1 29 0 . 138 ).

It dU be notlead that thoco findinGG onlato th000 of 3.2 9 a eoitont carHod out an a arass bround. Larre av=bom of biro uorn footr dm4ng 3.3 and. 3 . and It wao trpoe3blo to obtain data for oepwate cpoieo. Vat their rosponsoo uoo pvebably zolati1y hcz!ogenoo. Thio La p tio1avly noticeable In the case of 1G9D populatlanol , as ohatm by Toblo 29 and Table 23

all the bfr& UOZO 31tOt COp1OtO1y rootriated to ramovina biomo.

It fo1ioo that thQ hou OZG1O ) for ozatnplo ?Iat hco had prforoneoo

fbr bno In the two ezpexenta. This As contrary to other r3oltG (opointo 1 and 3) whom it iae noted that hooepazzcwa favzod greena. t posolble menan for this deCopacy and for the ualfamity of the it cLea predilecticno in Givon by the oboevation that the / bird had boon coon feed on e, ov In c pilec of bcmioh hop amwe near

the QaILtnntQ1 139 P 127)• Zf t7O accept the eLarj Imago a 2rlo%4,. thon the bfrC opy be opoctod to heo boccz c4iticaad to oeaehiarj for biccn pzay.

1lncidontally theco roaultc divoctly cctadict a otatoront by Cio (1967 .139) that 'a gmon mutmt v vogwdl000 of ftc 3 could not bldo Et pdatos on a hamaSancavo bvoun bac'ond'. - 120 -

TthXQ 21

awbor3 and porconteoo of baits taken on firet &ye of predation on 9C1B populations by b1cithfrde (B].) alone cd by b1eo1birdo rplw other birds (So m song hrushos > HS house oparrete, St starlings)

() In onparimonto whore 9G.-1B populations tore presented first,

opt. birds G B

1 B]. 63 9 12.5,

311 Bl,So,HS 45 19 29.7

3.4 B1,St 8HS 52 10 12.9

5.2 B]. 31 6 16.2

513 Bi 14 8 36.4

514 B]. 12 4 25.0

6.2 B]. $ 7 58.3

(b) In experiments where 9G: lB populations were presented second.

expt. birds G B

:2.1 B1,So 6 39 86.6

3.2 B]. 1 42 97.7

3.3 B1 e St eHS 29 17 36.9

4 B]. 32 20 38.5

5.1 - B]. 28 18 39.1

5.5 B]. 20 39 68.1

6.1 B]. 84 54 39.1

0 Dtci sire the total predation for the first 2 deyc - 121 -

(c) Con diti ni When these colourprefG!t3flCes wore taken into account by assuming that they were independent of the relative frequenCi35 of the baits offered, we found that, on average, selection was apparently apostatic in each experiment. This implies that the birds tended to hunt by searching images. Is there any direct evidence to support this assumption? Attention is focused, on predation by Tuz'dida, alone and with other birds,, after the change of populations in the IG: 9B + 9G: lB experiments. The second preliminary experiment showed that preferences for browns can be strong oven when these colours are rare • Tone of the repeat experiments produced such marked responses, though in 3.2 the results of the first to days after changeover disclosed almost

100% predation of browns (see Table 28, p. 136). In the remaining experiments the proportions of browns taken after changeover were generally highest during the first day. This initial overprodation of rare browns could possibly be due to natural preferences extrinsic to the experiments. If this wore the case, we would expect similar selection to occur at the beginnings of

the 9G:13 + 1(3:98 oxierimente. Table 21 compares the results from experiments where these populations were (a) and were not (b)

preceded by 1(3:98 populations. The proportions taken in the former situation tend to be higher than the values in the other set of data.

This suggests that the birds had 'carriedovor' preferences obtained

during exposure to 1(3:98 populations. In other words, their natural brown preferences had been strengthened: they were hunting by searching

imago. -122-

6. 9U2t4ARY

In 118 euperiments wild passerines on masse iere presented with populations containing one colour nine times as common as the other. To summarise, the main conclusions tere as follows if

A preliminary experiment revealed considerable heterogeneity

In the data and it was therefore necessary to Carry out

repeat studies. Blackbirds took more brctns than expected on the basis

of 9:1 ratios. In 13 out of 14 experiments the deviations from the expected,, based on the assumption of constant selection, wore in the

direction predicted by apoetatic selection. Taking into account the results for separate species, 19 out of a total of 20 sets of data gave evidence of apostatic

selection. (a) There is evidence that the birds were hunting by searching image. - 123 -

APPENDIX At TABLES 22-32

T rialobv-trial au1'ts fo 9:1 Meat exoriments of Ohatar 'II

(a) Key to Tables

The Tables give the following details of each experiment:

Species of predator and estimated uumbor of each (in parentheso)..

Type of background.

Composition of populations.

Dates and times of trials.

5 • Numbers of greens and browns taken per trial.

(rand totals of greens and browns taken from the tto types of population.

Expected grand totals, calculated on the basis of

constant selection (in parontheSese, below observed grand totals). 8 • Asterisks indicate unobserved trials.

9 • 's e indicates that over 50% of the area of the plot was covered by snow during the relevant trial.

(b) Tables and experiments/aee ovorleaf - 124 - Table 22 Experiment 3.1. Totals of baits taken during trials

Predators - blackbirds (8), songthrush (i), housesparrovis Background - grass - a. 90 greens:10 browns Date Times Predation (1968) (B.S.,) G B 1.8 12.15 - 17.00 30 9 * 17.15 - 20.00 15 10 * 2,8 9.10 10.15 8 10 * 10.20 - 14.30 18 10 * 14,45 - 16,30 14 - 9 * 16,52, - 20,31 5 10 * 3.8 9,02 - 11,05 73 8 * 11.31 - 17.05 .37 2 * 5,8 10,45 - 11,50 1 1 11,50 - 17.00 42 '10 * 17.20 - 19,20 9 7. 7.8 9,25 - 11,10 16 1 11,50 - 15,00 28 8 * 15.28 - 16,35 8 5 16.41 - 18,10 33 6 * 8,8 - 10,00 - 14,50 21 14 12,8 10,30 -. 12.00 5 0 Grand totals 363 120 (349.1) (133.9) b. 10 greens:90 browns Date Times Predation (1968) (B.s.T.) G B : 13.8 ' 11.39 - 12.33 1 19 12,40-. 15.45 0 44 * 16.03 - 19.41 0 42 * 14.8 20,10(13.8) - 9.01 0 79 * '9,20 - 14.01 0 18 '* Grand totals '1 202 (6.3) (196.7) - 125 -

(I) Presentation of 9G:lBpopu1atioiao followed by XGE9B populations

Experiment 3.1 t1etho4 The grid was laid down on a lawn of t4hden House, Newington,

Edinburgh (O.S. grid r3ferenco MT 249724) and populations %oZ'Q presented on the dates shown in Table 22 • ObCOI8od trials revealed that predation was mainly duo to blackbirds and a single C?) thrush, though house sparrows, dunnocko and starlinge were also involvod.

Of the 108 baits that wore observed being eaten, only 13 were consumed by the lost three species.

Results Table 22 gives the nuaboz'e of baits taken during each trial. Analysis 1 reveals significant heterogeneity in the proportions of greens and browns taken per trial in 3.1a (V. 5) 4. 0.0 fter changing the relative frequencies of the colours, the ecumon browns were taken almost exclusively on each day.

If the grand totals are compared with the expected 9g1 ratios assuming random predation, it is obvious that relatively too many browns were taken from both 3.1a and 3.lb • A measure of this overall brown preference is given by

C.P.R., assuming constant selection a 3.46

Taking this into account, we derive the expected predation. hoirn in parentheses in Table 22. The recorded selection deviates from the expected in the direction predicted by apostasy. This effect was statistically oiificants

6.6$0 P 4 0.02 - Ci)

1Tho 2 baits token in the first t4a]. on 5.8 were pooled with the data for the succeeding trial, whilst the 5 browns taken on 12.8 were added to the results of 8.8. - 126 -

Table 23 Experiment 3.4. Totals of baits taken during trials

Predators - blackbirds, starlings, housesparrows (many of each) Background - soil a. 90 greens: 10 browns Date Times Predation (1968) (B.s.T.) G , B 13.8 17.08 - 17.45 24 5 * • 21.17 - 21.32 28 5 * 14.8 10.16 - 10.38 29 8 * 12,21 - 12.30 21 10 * 15.15 - 15.29 12 10 * 15.8 17.01 - 17.11 14 9 * Grand totals 128 47 (112.5) (62,5) b, 10 greens: 90 browns Date • Times Predation (1968) (B,s.T.) G - B 16.8 13,01 - 14.15 1 31 14.15 - 14.31 0 48 * 14.51 - 15,10 0 53 * 15.16 - 15,30 0 12 * 17.8 10.01 - 12.09 0 86 * 12.15 - 13,59 0 32 * Grand totals 1 262 (4.7) (257.3) 127 -

imont 3 !!: flethod The experiment was carried out on a fallow plot (O.S. grid DGfGX'OnCe 275713) of a market-garden near Bridge End, Edinburgh. The green baits appeared more conspicuous against the soils Fox obvious reasons the area was woll..endowed with birds and the rate of bait-predation was correspondingly high. 1any birds were observed feeding on a pile of hop.imanure about 100 m from the plot.

R4EiSuitS Table 23 gives the results • Hhen gveit were common, the proportions taken per day were significantly heterogeneous

(X 5) 6.25 P 0.05). When browns became common they were taken almost exclusively by the birds. A comparison with expected 9*1 ratios, shows that too many browns were taken in both 3.4a and 3.b. In other words the birds posos5od fraquenoy4ndependent preferences

C.P.R. assuming constant selection a 5.00

The expected predation, assuming constant selection, is shown

In Table 23. In comparison, the recorded values deviate in the direction predicted by apoatatic selection:

X1) = 9.00 9 P ' 0.01 - 128

Table 24 Experiment 5,2. Totals of baits taken during daily trials

Predators - blackbirds (4) Background - grass a. 180 greens: 20 browns Date Times • Predation (1969) (G,M,T.) G B 14.2 10,01 - 13,50 31 6 * S 15,2 10.11 - 10,53 35 8 S 16.2 9,00 - 13.45 34 5 * S 17.2 12.05 - 17.08 47 4 * S 18.2 12.59 - 16.22 51 4 * S 19.2 11.55 - 15.12 43 9 * S . 21.2 9.45 - 15,20 50 8 * S 23.2 11.14 - 16.12 62 12 S Grand totals 353 56 (343.0) (66.0) b. 20 greens:180 browns Date Times Predation (1969) (G.11.T,) C B 24.2 9.00 - 12.05 3 51 S 25.2. 8.55 - 17.01 5 73 S 1.3 11.30 - 14,59 1 32 * S 2,3 9,10 - 16.50 o 49 * S 4.3 9.49 - 16,20 0 .21 * S * 7.3 9045 - 16.10 52 Grand totals 11 278 (17.4) (271.6) - 129 -

Enird.mlant - 5.2

ethod The grid was situated in a field (0.S. grid reference N? 29610) about one mile south of ROsewell, 11d1othian. Snow overed the site for meet of the operiment. Botucen 15th and 25th February LnclvaLva the plot was totally covered. On the 14th February and let, 2nd, 3rd and 4th March snow covered about tbreo4iartrs of the area • The thro days of continuous obDorvatian revealed that blackbirds were the only predators.

Results The results for daily predation ere ohom in Table 24. There is no evidence for heterogeneity in the nuber'o of greens and brone taken daily during 5.2ct (Xi, ) = 5047 P 0.50) • After the compositions of the populations tiara changed, the birds scan began to concentrate on searching for browns.

The blackbirds had an overall preference for breme:

C.P.R., assuming constant selection 0 1.73

The expected predation, assuming constant selection ,is sham In Table 24 and it can be seen that the recorded valves deviate in the directions predicted by epostatic selection. Theco doviaticn aria statistically siificant:

1.32, P < 0.05 i) - 130 -

Table 25 Experiment 5,3, Totals of baits taken during daily trials

Predators - blackbirds (2) Background - grass a. 180 greens:20 browns Date Times Predation (1969) (G,M.T,) G B 17.2 11.30 - 16.01 14 8 * 18,2 10,09 - 14.59 38 7 * 21.2 10,59 - 16.25 48 10 * * 23.2 . 9,40 - 16.59 51 8 27.2 12,15 - 16,52 31 2 * 20 10.25 - 16.30 49 5

- Grand totals 231 40 (217,0) (54,0) b, 20 greens:180 browns Date Times Predation (1969) (GJ'i,T.) 0 B 3.3 11.10 - 13.40 3 .33. 403 10.15 - 17,00 2 28 * 6.3 10.10 - 16.59 0 50 * 8.3 11.30 - 17.48 1 76 * 9.3 12.50 - 16.35 0 66 *

10.3 9.59 - 16.25 1 48 Grand totals 7 299 (4.5) (291.5)

0 - 131 -

!orirnent 5.3

Method The as sited in a field (0.S. grid reference NT 261702) on a north-facing elope 50 m from the Braid Burn in

Blackford Glen, Liberton, Edinburgh. Blackbirds were the only observed predators. The plot was completely covered by snow on

the first five days. During the next four, patches of grass were showing, whilst for the lest three days of the experiment the

ground was clear.

Results Table 25 shows the daily reulto. The proportions

taken during 5.3a were significantly heterogeneous (x 6) 11.82 9 P < 0.03) and during 5.3b the birds clearly tended to concentrate

on predating browns. Comparison with the expected 9:1 ratios shows that too many

browns were taken from both parts of the experiment and this

preference is illustrated bys

C.P.R., assuming Constant selection = 2.2

The recorded predation deviates from the expected values

calculated by assuming constant selection in the direction predicted

by apostatic selection:

x2 = 8.600 P < 0.01 (1) - 132 -

Table 26 Experiment 5.4 Totals of baits taken during daily trials

Predators - blackbirds (2) Background - grass a. 180 greens: 20 browns Date Times Predation (1969) (G.14.T.) G B 8.3 10000 - 16.19 12 4 * 9,3 10.58 - 11.50 18 5 10.3 9.07 - 15.41 29 3 * 7,4 - 10.05 - 17.28 31 8 * 8,4 11.01 - 14003 23 3 * 904 12.15 - 17.05 33 7* 11.4 9.56 - 14.04 46 9 Grand totals 192 39 (175.6) (55 0 4) b. 20 greens:180 browns Date Times Predation (1969) (G,I.T,) G. B 14.4 10.02 - 16,10 3 63 15.4 11.34 - 16.45 0 46 * 16.4 12.10 17,59 1 54 * 21.4 12,56 - 15.05 2 39 * 22.4 9.58 - 17.15 0 77 * 254 9.15 - 17.45 0 69 * Grand totals 6 348 (1306) (340,4) - 133

Eperimet

4ethod The grid was placed in a grassy field (0.S, grid reference NT 300625) near RoseweLt, Nidlothien, and blackbirds were the only observed predators. As can be seen from the 4atee given in Table

26, the experimrnt was suspended for about one month three days after the start.

Results Table 26 shows the daily predation • The proportions taken during the first part of the experiment did not differ significantly (X6) - 3.34 9 P > 0.70). Compared with 9:1 ratios, too many browns were taken from both types of population: C.P.R., assuming constant selection = 2.84

Taking this preference into account s the recorded predation deviated from the expected in the direction predicted by epoetatic

ectLon. This effect was statistically significant:

10.82, P 0.01 X( 1) 134 Table 27 Experiment 6.2. Totals of baits taken during trials

Predators - blackbirds (4) Background - grass

a, 180 greens:20 browns Date Times Predation (1969) G B 5.5 9.59 - 16.14 5 • 7 * 6.5 9.30 - 11.30 33 • • 13 * 9.5 9.30 - 11.30 36 9 10,5 14.02 - 16.30 42 7 • 16.30 - 18.00 25 •. • 8 * 1105 9.45 - 12.10 40 • 16 12.15 - 14.30 35, 9 * 15.01 - 17.45 .63 14 19.5 9059 - 10.20 .12 5 • 10.30 - 12.16 28 8 * 12.29-- 14.45 43 9 * 16.28 - 17.16 13 1 20,5 13,57 - 15,06 40 6 15.18 - 16,45 25 6 * 25,5 9.59 - 11,40 .43 , 5 - 12,00 - 14.35 15 2 * 27,5 13.30 - 16.25 48 29,5 9.09 12.12 33 6 30.5 14.12 - 15,10 18. 2 15.15 - 16.21 22 • 6 * Grand totals 619 • 148 (601.3) (165.7) b, 20 greens:180 browns Date . Times Predation (1969) G • B 4.6 12,58 - 17.07 5 67 6.6 11.39 - 12.10 0 21 12.20 - 14,39 o . 37 * 10,6 . 10,15 - 11.52 0 24 12015 - 14.59 1 .69* 15,03 - 15.04 0 5 15,27 - 17.31 0 90 .* Grand totals 5 313 (13.6) (304,4) - 135 -

Exjrinent 6.2

Method The experiment was carried out on a playIng field (O.S. grid reference NT 272704) at Liberton Dams, Edinburgh. Blackbirds were the only observed predator's.

Results These are given in Table 27. The proportions taken per trial are.not significantly heterogeneous ()(2 19 27.631, P < 0.05) during 6.2a. Following the change of population composition, there may have been a tendency for the birds initially to search for greens. On the first day of presentation of 20 greens: 180 browns, 5 greens and 67 browns were taken as opposed to a total of 1 green and 256 browns for the remaining six days.

These proportions are significantly different (X1) : 13.493, P <

0.001). Moreover, all 5 greens were taken during the first six visits out of a total of 15 for the day. Overall, however, the birds took too many browns from both types of populations:

CSP.R., assuming constant selection, 2.48

Taking this into account, the birds tended to select in an apostatic manner and this effect was significant:

X(1) = 8.13, P < 0.01 - 136 -

Table 28 Experiment 3.2. Totals of baits taken during trials

Predators - blackbirds (4) Background - grass a. 10 greens: 90 browns Date Times Predation (1968) (B,s.T.) G 1.8 1140 - 16.15 0 • 45 * 17,12 .- 20,45 - 0 71 * 2.8 21,13(1.8) - 11000 1 90 * 11,28 - 15.00 0 44 * 15,20 - 17.00 0 44 * 17.19 - 20.15 0 88 * 3.8 20,43(2.8) - 12.45 0 90 * 13.11 - 14.48 1 82 * 15,18-. 20.00 0 90 * 4.8 20,32(3.8) - 12,00 0 90* Grand totals 2 734 (4.0) (732.0) b. 90 greens:10 browns Date Times Predation (1969) (B,s.T,) G B 4.8 15.40 - 16.22 1 42 6.8 13,40 - 16,30 0 28 9.8 11.00 - 12.40 .17 13 * 12,8 14.35 - 16,30 5 2 16042 - 19007 30 7 *• 13,8 9,02 - 10.01 10 .10 * Grand totals 63 102 (5005) (114,5) - 137 - (ii) Presentation of 1G:93 pqpulatione folloted ky 1G,9B ppulatious

Experiment 3.2

2ethod The grid was situated in a disused part of a playing field (0.8o grid reference NT 272704) at Liberton Dams, Edinburgh.

The grass at this site was more uniform in length and colour, and more dense than in the agricultural fields used in other experiments

The only observed predators were blackbirds.

Results The numbers of baits removed during each trial are shown in Table 28 • Browns were taken almost exclusively during 3,2a* On four occasions this colour was completely eradicated from the population and in only one of those cases were any greens taken - and then no more than one. The daily proportions removed from the 9G:1B populations were significantly heterogeneous (X)2 t 79 87, P c 0.001). During the first two days the birds took almost only browns, whilst later they ate relatively more greens. Overall,, the birds took too many browns from both typos of population:

O.P.R., assuming constant selection = 20.40

Taking this into account, the birds tended to predate in an

postatic manner. This effect was not, however, statistically significant:

54!44, P 4 0.02 - 138 - Table 29 Experiment 3,3. Totals of baits taken during trials

Predators - blackbirds, starlings, housesparrows (many of each) Background - soil a. 10 greens:90 browns Date Times Predation (1968) (B.s,T.) G B 3.8 12.00 - 2 1030 0 40 * 4.8 21057(3.8) - 12.32 2 52 * 12.48 - 15.00 0 15 * 58 12.00 - 14.25 0 90 * 7.8 9.01 - 14015 0 90 * Grand totals 2 287 (5.8) (283.2) b0 90 greens:lO browns Date Times . Predation (1968) (B.s.T.) G B 13.8 14.05 - 14059 15 9 * 17,01 - 17.30 14 8 * - 20.50 - 21.15 8 7 .' 1408 9059 10.11 9 10 * 10.20 - 11,51 1 10 * 15.8 12.15 - 16,33 22 10 * 17.01 - 18.01 . 24 9 * 1608 9.09 - 9.55 10 9 * 10001 - 10,21 8 8 * 10.30 - .10.45 1 10 * 10.59 - 11.19 14 7 * 11.30.- 11.58 8 10 * 15.40 - 16,05 9 10 * 16.51 - 17.08 11 10 * 18.21 - 18.05 7 8 * 18.10 - 18.45 6 4 * Grand totals .167 139 (159.1) (146.9) - 139 -

Exper .3.3

Notho4 The grid was laid down on soil, in a separate part (O.S, grid reference NT 277713) of the market garden used in 3+ The two plots were separated by a main mad and a di3tenco of about

mile. Many birds of various species wore involvod, in the experiment 3.3 and they were also seen feeding in a pile of hop-manure about 30 m from the grid.

Results Those are given in Table 29 • The birds clearly had strong preferences for browns. In 3.3a the common browns were completely eliminated on two occasions. The proportions taken per trial during 3.3bere distinctly heterogeneous (X 15 ) 30.16 0

P 4 0.02. Overall, too many browns were taken from both typos of population; C.P.R., assuming constant selection 841

If this is taken into account s we find that the birds were tending to select in a frequency-dependent manner though s, statistically, this affect was not significant:

x 1) 3.035 0 P > 0.010 - 140 -

Table 30 Experiment 5.1. Totals of baits taken during trials

Predators - blackbirds (4) Background - grass a. 20 greens:180 browns Date Times Predation (1969) (G.M.T.) G B 14.2 11.09 - 14.45 .0 9 * s 15.2 9.10 - 15.29 0 10 * 16.09 - 16,29 0 11 16.2 11.08 - 16,00 0 35 * s 17,2 10000 - 15.11 1 38 * 18,2 11.22 - 15.49 o 68 * 19.2 12.50 - 16045 o 45. * s 21.2 12.00 - 17010 1 69* s 25,2 9,21 - 17,12 1 83 s Grand totals 3 368 (13.5) (357.5) b. 180 greens:20 browns Date Times Predation (1969) (G,I.T.) G B 26.2 12.45 - 16.36 28 18 s 28.3 10,05 - 13.49 42 16 s 4,3 11.25 - 17.45 59 15 * s

5.3 12.15 - 17,32 30 12 7.3 11.00 - 17,38 78 15 * 8.3 10,59 - 17.10 63 9 * 10.3 9,50 - 1605 72 6 * Grand totals 372 90. (348.0) (114.0) - 141 -

_ent 9±4

Method The operi2ent uan crriod cut in a grassy field (0.S. grid

ofoznco N% 30337) at Dalhousie Choctom o near RoSeTioll s Iid1othian. Sueti completely coved the cite during 5.1a and tho firot two &y

of 5.1b. On the third day only a few patches of enmy tozo left and by the !text day it had cosplotely tnoltcd. Blackbirds we.= apparently the only prodetorz.

Resulte The predation during trials is gizrn in Table 20. As with Other esperimente s the proportions of broms prodated vera

V3Er high when those bzit wore coon. Mon groena were

In 5.lb, the daily proportions taken were bet'ognoous (1 6 ) 23.210 P c 0.003.) • In addition, there was a ciif'icant increase in the perconta3 of grease token over the duration of S .lb (Spearean'c

rank correlation coefficient m 0.0329, P < 0.05). The grand totals abca that relatively too many browns wore rood from both types of population:

C.P.R., aaagtt constant selection o 2.95

!bon this weroU bran preferenes is taken into conaidoaticn It is 2oi4 that the rccorod predation deviates from the expected in the direction predicted by apostatio selection:

Z 1 ) 15.10 9 P < 0.003 142 -

Table 31 Experiment 5.5 Totals of baits taken during daily trials

Predators -blackbirds (5) Background - grass a. 20 greens:180 browns Date Times Predation (1969) (ô.I,T.) G B 7.4 9.10 - 16.49 . 0 35 * 8.4 12.00 - 17,15 0. 48 * 904 13.11 - 17.50 0 46 * 11.4 15,02 - 17.08 0 22 * 14.4 9.10 - 16.40 2 59 * 15.4 12.20 - 16.10 0 89 * 16.4 13.08 - 17.10 1 51 Grand totals 3 350 (14.2) (338.8) b. 180 greens:20 browns Date Times Predation (1969) (G.M.T.) G B 17.4 13.01 - 17.05 20 39 1804 11.09 - 16005 58 27 .1904 9.30 - 17.10 . 79 13 21.4 13.40 - 16,07 - 21 5 * 2204 11.10 - 17,50 83 3 * 23.4 12,55 - 17.15 40 5 * 24.4 10.00 - 18.10 60 8 Grand totals 380 81 (356.1) (104.9) - 143 -

Experiicent 5.5 .

Jethod The grid was situated on a lawn (O.S. arid *efezancs 1T 268706) adjacent to the Doparttiiont of Engineering, Kings

ni1dingu, Uniez'sity of Edinburgh. BlaCkbirds were the only apparent predators.

Res1tø The 4tLtly numbernumbera of baits taken are given in Table U. After an almost monotonous diet of browns in S .5a, the birds also took relatively too: many of this colour when rare in 5.5b. The proportions taken daily were ainificantiy heterognooue (X 6 ) 100.454, P 0.001).

A m3asure of the brozn preferences is given by:

C.P.R. 1 assuming constant selection e 2.65

Taking this into account,, the birds tended to predate in an apostatic mz nor:

X1) 16.24 6 P c 0.001 - 144 Table 32 Experiment 6,1, Totals of baits taken during trials

Predators - blackbirds (4) Background - grass a. 20 greens: 180 browns Date Times • Predation (1969) G B 7,5 14.15 - 15,20 o 18 15,30 - 17.29 1 16 * 8.5 14.20 - 15.10 0 23 15.15 - 18.11 1 55 * 12.5 11.15 - 11.29 0 6 11.30- 18.41 1 110 * 13.5 9.10 - 9.12 o 9 9.16 - 12.00 0 85 * 18.5 10,37 - 13.49 o 56 * 14,09 - 14,10 o 2 14.15 - 18.05 1 89 * 22.5 9059 - 11.39 0 35 11.42 - 12.00 0 50 * 25.5 14.10 - 17,00 o 52 Grand totals 4 606 (593.6) b. 180 greens:20 browns Date Times Predation (1969) G B 26.5 10.15 - 17,15 84 54 28.5 14.00 - 15.14 7 12 15.16 - 17.18 37 11 * 2.6 14.15 - 14.45 7 2 14.52 - 17,20 70 10 * 3.6 10,30 - 12.01 29 12 12.15 - 14.10 33 5 * 5,6 14.21 - 18.59 .46 9 * 9,6 15.19 - 18.49 41 7 * 13.6 10.09 - 17.15 59 9 * Grand totals 411 131 (37504) (166.6) - 145 -

Exjeriment 61

Method The experiment took place in a grasoy field (0. S. grid reference NT 302618) near Capie1ati Farm * Rosow11 0 Piidlothian.

Again g blackbirds were apparently the only pz'edatorn.

Results The numbers taken daring each trial are shown in Table

32. When brotns taz'e common predation was almost exclusive to this colour. During 6.lbf, when groans were common s the proportions taken per trial were ignificaflt1y hetoroeneous(x 9) = 50.50w

P c 0.001). Note the Mob percentage of brosns taken on the first day (26thNay)* A uraasure of the brown preferences is given bye.

C.P.R. 1 assuming constant selection a 3.99

Taking this into acgOt 1 the recorded predation dozinted from the ,expected In the direction predicted by apostatic selection.

This effect was atatiatically si4ficants

X1) 2O.?2 p < 0,001 - 146 -

VII POISS OF 111DIVIDUAL ''URDIDAE - 1'47 -

Figure 10 aAL VII Sites of plOts in experiments 2.1 (Chapters91 and 2.2, 2.3 (Chapter VIII )' , showiiig boundaries of bla ckbird territories a. Experiment 2.1 (January - February 1967)

fb ) 0

L__--- f • ] KEY / Drive ----_ outline at EJ (fb plot. s—bush tree b.. Experiments 2.2 (March - April. 1967) and TV ii1JVYi L/(4 2,.3 (January - February 1968) law n -.alf( long grass 0 territorial I * - boundary ~Z-f~ b observation i L- — Point

Sc/e / 1 L 0 5 10 Metres 1

1L/Ct/SE "

(Adapted from Fig. 11, Allen 1967) - 148 -

CHAPTER VII M IMSPOINSES OF XDI VI DU1L TMIDAZI PART 1

1, INTRODUCTION

The results of the first 9:1 oeient (Chapter VI, this thosiag ale 0 AUoa 1967 6 Allen and Clarke 1968) z'ovoalod hatoroffmofty batmen the pzoportiono taken by blackbirds per visit • This effect may havo been a rocult of diffonc3s bettoen the bohAvioiw oZ individual birds, but this possibility was mteetabla bocaue the birds could not be relicbly identified. In the epericxnta dancribod in thie and the foUoing Chapter all the L!Ldus individaa.b werm co1ouruinpd. Three sets of emporimnts (2.1 0 2.2 1, 2.) were co1atod at the eeae io • Thin Chapter fit dacribeo the general detii10 of protedure cd the ecu1te of expericent 2.1 am then preeeted and dinoueeed.

2. EXPflITAL AREA ID BASIC 11AWRIALS AND METHODS

(a) eritauta1 dwo

The oxpoz'icante tier caiod out on a Xan of riiy Reading hem (O.S. grid zvforencc SLI 71676). Figure 10 ehce the general topography of the aroi. The lean cneieto of about sixty per cent greoc and forty per cent daisy (I3o]1ie perennig L.) and plantain

(Plantai eadia L.) • A ekewez'..uax'ked grid of 100 etre.euazc tea placed on the 1cn in approdactoly the oaro position in each expeicent. In experiment 2.1 it was equareg in the subsequent studios it wco L'..chcped. 0ooratione tore made from behind covr In a porch of the hotac • The populations tcro maintained in the norit =naor (coo Chapter XII) and data wore ocordod ancietly an outlined on pegos 60-61. - 149 -

Results were obtained for each individual blackbird and songthrush and for predation by groups of dunnockc3, robins and house sparrcea (see below) • The populations were continuously observed and they ware usually presented from dawn to dusk.

Hence a complete history of behaviour was built up for each bird.

Particular attention was paid to 'imprint-pecking'. This term describes the habit of some blackbirds and thrushes to pick up a bait and then drop it • The result of this action is a bait imprinted with the bird's beak hence 'imprint pecking'. The term is simply one of convenience: it describes the act, not its cause.

(b) Bird The blackbirds frequently showed territorial behaviour and from the positions of their conflicts it was possible to estimate the approximate boundaries of their territories • Those are shown in Figures lOs and lOb. An important point to notice is that most or all of the plot in each experiment was under the rule of a single pair of blackbirds • All intruding birds were liable to be put to chase. This particularly applied to blackbirds and, to a lesser extent, to oangtbrushes. The interlopers had little time to choose their prey when they were under attack. Colour-ringing of the Turdidae was carried out at various times, but most of the birds were ringed before the starts of the experiments (see Alien 1967, Table XIV) • The majority of the birds were caught in a hand-made 'lobster-pot' trap baited with

'Swoop and/or bread. Dunnocks, robins and house sparrows were also coXour.,rind but though some of these birds were observed - 150 -

feeding on the baits, accurate recoition of individuals was

difficult. This was mainly because of the relative sizes of their rings and the swiftness of their feeding (particularly with

regard to robins).

TemperqtgM, The teimporature at noon was recorded for every experimental day. During experiment 2.1 (January February

1967) the weather was exceptionally Dud for the time of year and a mean of 9.6 °C was recorded. For 2.2 (I1arch April 1967) the aen mid-day temperature was 10.3 0C and for 2.3 (January - February 1968), 6,70C.

Avaiab1e natural food. Despite the relatively high average air-temperatures recorded during the two winter experiments, the ground often remained frozen • Worms, the main diet of lawn- feeding Turdidao, were therefore inacceesiblo • It would seem that the birds relied on baits as a main source of food, for rarely were they observed feeding otherwise. In contrast, the birds of 2.2 (particularly the 'resident' pair) frequently entered the plot and ignored the baits, preferentially hunting for worms • No accurate assessment was made of the uwnbere of worms taken.

Rein. The work was never curtailed by heavy rain; observations ware always continued, although the birds tended to visit the plot 1088 frequently. Snow was never observed to eettle though on some early eore.thge the ground was covered by frost. - 151 -

(IV) Gieeo • In the winter- epor'itente the greec was not groiug but was 'long'. On avorqp o the b1doo mra probebly about 3 cm long, end of conrea not all tere vwticel. Four &iye before the start of 2.2 the lawn was crn and by the end of the onporimnt the grass had grown to the length it had been in 2.1.

S. ZXPXEUT 2.1 1 10:913 P0PULA?I0ITS F0LLOUD BY 9G: 1B ID OTHER 10 -D0TIi7àTED' POPULATIONS

This otudy hee been described briefly in Chapter VI in

Ccin with the overall predation • Hove we concidor the birde individually. Two blackbirds and a congthruah wore pruentod with 10:98 populations eteocoded by 9s1B and other typos of populations containing succeaeicety fewer: freqt!onoioo of br'owne.

The dates of preeentatin of the popu1ticns wore as follows From 16 to 23 January (iucluiive) : 20 greene 180 brotmo From 24 January to 31 January (inclusive) :190 greene + 20 browns. a. Onl and 2February 4. On S February 110 greens + 14 browno e. On 4 Fobrwiry :199 greene s 1 brown

The reesoning behind the presentation of the ouceecoively lower frequencies of brown will be clarified shortly.

(b) Materials end .ethodo Figure ba ohows the position of the grid situated within the territory of a male colouringod blackbird (A) and his mate (13).

Thio o*pevisnt forced port of a B .Sc. Honours project (Allan 1967) end is inc1u!ad hero for epbotoneoe. -152-

Table 33

Experiment 2.1. Grand totals of baits eaten and imprint-pecked (shown in parentheses) by blackbirds A ( e ), B ( ) and songthrush T (on).

LA.CKBIRDS TIRUSH Inclusive dates Population of presentation A B T (1967) 0 B 0 B G B a. 200 + 180B 16.1 - 23.1 0 125 1 122 0 65 - - (2) (0) - -

b, 1800 + 20B 24.1 - 31.1 1 130 91 21 0 176 (11) (o) (12) (o) - -

1920 + 8B 1.2 - 2.2 0 30 32 0 0 33

(2) (0) .- - - -

1960 + 4B 3.2 15 2 12 0 0 21

(29) (0) .- - - -

1990 + lB 4.2 20 0 20 0 11. 0 (1) (a) - - () (a) - 153 -

These birds along with a itialo thrush (T), were the maim predators.

(a) Results

Ci) Summary-of data Table 33 condenses the results for the individucU. Turdtdse 1 .

The complete daily results are given in Table 36 in Appendix B e'

-page 161(frdm Allen 1967 1, Table IV) • The plain figures represent the numbers of baits consumed by each bird over the periods of presentation.

Figures in parentheses represent the grand totals of imprint-peeked

baits over the corresponding periods. The birds clearly differed in their individual responses. Although they all took an excess of browns from the 1Gg9B populations, only one bird, B, took appreciable proportions of greens from -9G: lB populations. Blackbird A and the thrush T consumed browns almost exclusively. It ww for this reason that the relative proportions of browns were subsequently lowered. In other words the birds' preferences for browns had led to the question: how low must be the frequoncy of browns before A and T are ,forcod to eat greens? Or more amply, how 'strong' are their brown preferences? As can be deduced from Table 33 2, A switched to feeding on greens when there were only 4 browns within the plot whilst T did not relax his brown preference until only 1 brown was available • It will also be noted that the relative incidence of green imprint-pecking in these birds was highest on the days when the behavioural switches occurred. Of B's 12 green imprint'.pecks, 4 occurred nn the day when she first stamed to take the coon greens (24.1.67) and 5 took place

on the second (see Table 36, page 161). We shall now examine the individual responses in greater detail.

tRobins and dunnocks took greens alone from the four types of green- dominated population (79 greens and 9 greens, respectively). Neither species removed any baits from the iGi 98 populations. - 154 -

(Li) pE3t8tiC OQ1øCtiQ!

ie noted in Chtar 'II (page 108) that oz'aU1 the birds did not exhibit aignificant uWay-dependent osleotion. It us now exantno the results for parts a, and b. with regard to the data for Individuals. On the basis of the 9:1 ratics it is clear that all the birds took too many browns from both typos of populations. A and 'F were obviously not preying apo8tatical1, for in both parts of the experiment they ate broto almost without exception. B on the other hand, took

,appreciable numbers of greens during the presentation of 180G 4 20B. Her results can therefore be tested for frequsncy..dependonco, using the method described on page iou. Once again, however, we are faced with the problem of heterogeneity within, the proportions taken per day. This point is made by the relevant data shown in Table 36

(page 161) 0 though these are not amenable to analysis by the chi- squared test. Nevertheless, if we assume selection to be constant In parts a. and b.. of the experiment, then we can derive the expected values for the grand totals as shown in Table 34. These differ from the expected

in the direction predicted by apostatic selection and this offset was statistically significant (x1) 5.45 P 4 0.02). The female blackbird appears to have been predating in an apostatic manner.

Table 34 Observed and expected (assuming constant selection) grand totals of baits taken by blackbird B during parts a. and b. of experiment 2.1 • The expected values ar's shown in parentheses. S B a. 1 122 (4.24) (117.76) 91 21 (83.24) (28.76) Tables 5a,- b. c

Diminution of brown-preferences in blackbirds A, B and songthrush T. Sequences of baits eaten and imprint-pecked per visit during the transitional periods. Visits' are numbered from the start of presentation of 20 greens + 180 browns. Imprint-pecked, baits are shown in ordinary parentheses. Symbols within square brackets pertain to the same bait. Thus on his 97th visit after 'ratio-reversal', T imprint-pecked a green before picking it up and eating it.

Table 35a Blackbird B

24.1.67 25.1.67 Visit Order of selection Visit Order of selection 1 2 3 1 2 3 1 b 14 g 2 b 15 (g) g 3 b 16 g 4 (g) b 17 g 5 b 18 g g 6 b 19 g '7 (g) (g)' 20 (g) 8 g 21' b 9 g g g 22 g 10 b 23 g (g) 11 g 24 g g g 12 b 25 g g 13 (g) g 26 g g g

Table 35b Blackbird A (3.2.67) Visit Order of delection

1 2 3 4 5 6 7 ' 8 9 10 11 12 13 14 15 16 17 97 (g) b 98 b

100 (g) (g) (g) (g) (g) (g) (g) (g)(g) [ () g](g) gJ 101 [(g) (g) g] (g) 102 (g) (g) g g 103 g g 104 (g) (g) (g) g g 105 (g) g g 106 g g

Table '35c Thrush T (4.2.67)

Visit Order of selection 1 23456

93 . nothing taken

94 II

95 ft tI

96 " 97 [(g) g] 98 [(g) gJ 99 g) g1g g'g g 100 9 9 gg - 156 -

(iii) Rola,wtion of brøwn preferences

As already eantioned the birds did not itediato1y stop searching for browns tn these baits became rare • Blackbird B was the only predator to switch to searching for greens as well as brne on the firet day. The evolution of this change is shown in Table 35a which gives the sequences of baits taken per visit during 24 and 25 January It will be noted that the first three greens solected (on visits 4 and 7) were not eaten, but merely piched up and dropped (imprint.' pecked). Blackbird A and the aongthrush continued to search for browns. The blackbird in fact imprint-pecked a total of U baits during the presentation of 180 greens + 20 browneg three of these occasions were on the first day after ratio reversal (214.1.67) and five occurred six days later (29.1.67). On his twelfth visit (25.1.67) he actually ate one green bait without hesitation (see Table 36). The thrush meanwhile appeared to disregard the greens entirely. As the relative proportions of browns were lowered, A was the first of the two male birds to lose his obvious brown preference. This took place on 3.2.67, when the population contained 196 greens + 4 browns • Table 35b gives the results for A on this day and shows the sequences in which the baits were eaten and isprint.pecked on each of the ten viito • As can be seen, the bird ate his first greens (apart from the single instance mentioned above) on the fourth visit. Flom then on, greens wore taken freely. This transition was not clear-cut, for the predation of greens was preceded by a relatively high inci&mco of green-imprint-packing. Such behaviour involved either different baits (as on visit 99), or the earns one (e.g. visit 100). On the succeeding final day of - 157 - the operiment, A imprint-pecked a single green and at- 20 more without hesitation. The male thrush totally neglected the greens until the last days when the population consisted of 199 greens + 1 brown. Table 35c illustrates the behaviour of the bird on this day. On his first four visits be searched all over the plot (see Fig. 14(b) Allen 1967) without finding the lone brown bait • Novez'theless he discovered about three minute pieces of brawn dough that had probably fallen to the ground during 'replacing'. Once again the onset of green predation was accompanied by imprint-pecking. The first three greens eaten were all imprint-pecked once before consumption.

(d) Discssion

He remarked in Chapter Vt (p,/OZ ) that the three birds as a group 4id not select in a frequency-dependent manner and, assuming homogeneous behaviour, it could be inferred that the same conclusion applies to the birds as individuals • This is refuted by a closer examination of the data, for One blackbird (B) tended to take the commoner coloure. On the other hand blackbird A and oongthrush T were strongly biased to take brown baits alone and irrespective of their freueny. Even those two birds were not alike in their behaviour; A started taking greens when four browns were present while T did not start until the population contained only one brown and then not until after' be had unsuccessfully searched the plot on four' separate occasions • The behaviour' of the birds therefore ranged from B to A to T, in order of increasing strength of their brown preferences. It may be relevant that B. the only bird to start eating appreciable numbers of browns on the first day in 2.lb, was the only- bird that had had experience of greens in 2 .la. - 158 -

The tho birds therefore d foiod in their rsc3ponsca. They also had other idicoynarwieeg for Inatanco with ragard to Imprint.

pokin, pariu1az1y when they were starting to food on aroma in This can be Goon from a comparison between Tables 35a, b, C. Othev differoncee of behaviour were loss moesurable. For oiamplo,b2ackbird A appoarod to exhibit a forz of Imprint-packing In addiion to the normal type • On those occeaicma the bird would crouch, bond Ito neck and oinply nip the side of the bait as it lay on the ground. This behaviour was recorded only on 29.1.67, thon At occurred five tko (see also Allen 19(37) • Blackbird B nevnr behaved in such a mennor and T but once, in experiment 2.2 (see Table 42, p.195 ). It will be noted that in two birds the incidence of green imprint-pocking increased markedly Ju3t before their 31Ltoh to gveen.feoding. Uo shell further consider this point later (9,183. if). All the other 9:1 G;qpr1=nta described in Chapter VI gave evidence that predation by groups of birds tends to be apostatic. I suggot that the present results vem due to the small sample of predators Involved, vhIch by chance happened to contain two iividuals with very strong preferoncoo for browns • The affect of these birds masked the effect of the tendency of B to choose the Cocnor oorto • The argunt that the present Rndin4p are the oxcoptiom rather then the rule is supported by the fact that the other thirteen ozperieente oil gave results in the diroction predicted by apoetatio selection.

Further discussion of the brown preferences and the variability of behaviour will be delayed until the and of Chapter VIII, after the results of eerieents 2.2 and 2.3 have been presented. - 159 -

(e) Conclusion

This experiment raised soma interesting qatLons for earnp1os

Ci) !Yh, do blackbirds and songthz'ushes prefer browns, If such behaviour is widopread?

To what extent were the strong bxottn preferences in 2.1b,

a and d due to searching images acquirad during exposure to the 9G1B populations of 2J.a?

IS the behaviour within a group of Tuxdus individuals always as heterogeneous as in the group in experiment 2.1? If so, what is the cause of the variability?

A control experiment starting with the presentation of 9t3 lB populations was required to help answer these questions. - 160 -

APPENDIX L3 i TABLE 36

'uil results for eporint 2 .1. Qtatex! VII -161- Table 35 ki Experiment 2.1. Full daily results for blackbirds A, B and Songthrush T. Imprint-pecked baits are shown in parentheses. The number of visits (v) per day are also given. BLACKBIRDS . THRUSH Date A 11 B. T (1967) G Btv G B I v G B I v a, 20 greens + 180 browns

1r7 •1 - - - - - 18.1 . , 0 24 11 1 34 15 - - 19.1 0 19 10 ( 2) + 0. 30 18 - - 20.1 0 16 10 0 16. 10 '0 6 3 21.1 0 21 12 0 15 13 0 18 9 22,1 0 24 14 0 15 11 0 19 7 23.1 0 21 10 0 12 11 0 22 8 Grand totals 0 125 67 ( 2) + 1 122 78 0 65 27 II b, 180 greens + 20 browns

24.1 (3)+o. 18 9 (4)+ 6 8 13 0 13 8 25.1 1 16 9 (5)+17 114 023 9 26,1 0 16 11 ( 1) + 17 1 11 0 20 9 27,1 0 96 5 140237 28.1 ( i) + 0 15 10 ( 1) + 9 .1 4 0 29 10 29.1 ( 6)+ 0 14 8 12 0 4 0 22 6 30.1 0 18 10 16 3 11 0 29 8 31.1. ( i) + 0 24 12 ( 1) + 9 6 10 0 17 5 Grand totals (11) + 1 130 75 (12) + 91 21 71 0 176 62

c, 192 greens + 8 browns

1.2 0 1710 15 09 016 7 2.2 ( 2) + 0 13 10 17 0 8 0 17 10 Grand totals ( 2) + 0 30 20 32 0 17 0 33 17

196 gre(ETns + 4 browns

3.2 (29) +15 2 10 12 01 7 11 0 21113

199 greens + 1 brown 4.2 (l)+2o 011 20 0 1 9 (3)010 + 11 -i&1-

CHAPTER VIII THE RESPONSES OF INDIVIDUAL TURDIDAE:

PART 2 - 160 -

CHAPTER VIII •HE RESPONSES OF INDIVIDUAL TURDIDAE: PPRT 2

1, INTRODUCTION

Epezisntc 2.2 cmd 2 . 3b wero essentially similar to the

previous study except that they started with the presentation of

9Gs lB populations. These two sets of experiments are considered

eimulteneously. In experints 2 .3c b dse an extra morph was added

to the a3rsteD end this section of the work is presented separately. The experitents1 site and the general materials and methods were described in the previous Chapter.

2, EXPERIMENT 2.21$ 9G:IB POPULATIONS.. EXPERXNENT 20 (PARTS ab)s 9G:18 POPULATIONS FOLLOWED BY IG:9B POPULATIONS

These studies were of comparable desigm and involved the same

Turdidae with the exception of one individual.

chronology The dates of presentation of the populations wore as follows.

Et. 2.2, From 20th March to 3rd April 1967 (inciusive)t 180 greens + 20 bins Expt. 2.30. From 13th January to 31st January 1968 (inclusive): 180 greens + 20 browns

b • let February and 2nd February 1968: 20 greens s 180 browns

(for c3d,o see page 175).

Materials en4 methods

Eeimen' 2.2. By 2arch 1967 9 the local blackbird population had

changed drastic&.ly (Figure 10b P. 147) though sangthrosh 7 was still in the vicinity. Blackbirds A and B had disappeared from the

1This oerLment formed part of a B.So. Hcnoure project (Allen 1967). Table 37

Exerimentsp 2.2 and 2.3a,b. Grand totals of baits eaten and imprint-pecked .(shown in parentheses) by blackbirds C (v), D ), E (o) G () and songthrush T Expected predation, on the basis of 9:1 ratios is also shown for each bird.

Incl. dates B L A C K B I R D S TIffiUSIi Popn. of present- C D E G T ation. G B 0 B G B G B 0 B 2.2 1800 - 20.3.67 97 113 140 57 158 6 24 5 7 117 + - 3.4.67 (184) (151) (30) (5) (32) (1) (1) (2) (6) (i) 20B

9:1 189.0 21.0 177.3 19.7 147.6 16.4 26.1 2.9 111.6 12.4

0) 2.3a. 1800 16.1,68 101 21 178 59 72 15 153 16 - - + -31.1.68 (58) (19) (io) (6) (3) (i) (7) (4) - - 20B 9:1 109.8 12.2 213.3 23.7 78.4 8.7 152.1 16.9 - -

b. 20G 1..68 0 21 1 27 1 16 4 43 - - + -2.2.68 () (22) ------- 180B

1:9 2,1 18.9 2.8 25.2 1.7 15.3 4.7 42,3 - - - 165 -

eea and n00T3 C (male) and D C) patr'o11d Mot of the zid. At the etct of the e oient the arid was the cc end chepo m in oW rI=mt 2.3. On the third day Ci nle blackbird E periodically attempted to feed on baits in the e astorn pozi129t0z' but was invariably chased =ay by one or both of the Irwidont o paL. The shape of the plot uan thezrofozo e1tozed cc that this bird vould have Smator opportunity of feeding on the baits. The con3thz'uth tics also s ubject to attack from C or De as tiec mother c1e blackbird (G) which ra gaarly onteod the grid 2oti a tritczy situated to the north (Fig. lob) • brnnooIcs fed on the pop1aticno tcade the end of the experim ent.

Emorimmt 2 • 3ab • Ten rzonths later the b2.acbir4s still eppoeed to be occupying the se tozitoies, but T had since left the cze. zcbino and house epavzvs also fed on the baits.

(i) sy~mnt of

Table 37 gives the grand totals of baits eaten and i.intpocled durints eerinxbnto 2.2 and 2.3ab. The e xpected grand totals of baits ea"a q assunins 9*1 ratios arc also ohctn. The data refe r to the five cola zind Tuzdidao only: the conpleto daily results for these end other bfrda (cbine 4tnocke and horse cparxtis) arc ohotin in Table 42 6 po 195 (fron Allen 1967, Table VIII) end Table

439 n3ge 196. In the follotiing pages attention will be eoncentrctod eainly on experiments 2.2 and 2 • Se. The results for 2.3b icply that all the ratoved are CM12. end do not c1le7 eeeningful concluiono. - 168 -

(ii) Predation

Ile first examine the proportions of greens end.bz'ome that wore actually eaten. Our interest centres around to main qsstions. First do different birds tend to take oluiler proportions, or othezioo? Second, doo the ears bird tend to take similar proportions from, two near-identical ezperitaente separated by cm Interval of tenmonths? But before these qsetions oz'o cmewered, It is necessary to establish whether any surcos of heteroganeity are being concealed by the overall results shown in Table 57,

Eerint2.2

Retezneitybetvoon &e • Statistical analysis for hoterogoneity is haipered by the fact that each bird took low numbore of baits per day. Nevertheless an examination of the data shown in Table 42 does appear to reveal such heterogeneity in at least three birds. For example, all 6 of the browns that word eaten by E were taken during the last two days of the exporicent.- C appeared to oat an appreciable majority of brome on sore days and of greens on others. Indeed the data suggest that this bird took progreesivol3r higher percentages of greens as the expericent proceeded (see also Allen 1967 9 Fig. iSo) • This bird's main change of behaviour appeared to occur on the tenth day (29.3.67). Previous to this

tica, the percentages of browns eaten had never reached 50% 9 whilst afterwards the percentage predation o.ays exceeded 37%. If we pool the daily results within each of these two periods we obtain the data shown in Table 30 • The two proportions of eaten baits are clearly different.

- 167 -

Tthe 38

2aie blackbird C. Numbers of baits eaten and imprint-pocked (in parentheses) during the fit 9 days and last 5 days of experiment 2.2.

200.67 - 29.3.67 27 100 (90) (129)

20.3.67 3.4,67 70 13 (9e) (22)

D. like her mate, also C nancod by qoncentrstin.g on bzotms.

Thouj çpoaring to adapt quickly to a green-biased diet (on the second day) he occasionally seem3d to relapse into taking

relatively higher proportions of brotns (e.g. on 24.3.67 and 30.3.67),

Overall z'osonses. It is clear from Table 37 that the Turdidas differed in their individual ovoral3. responses.'Three of the

birds (C, D, T) ato relatively far more browns than expected on the basis of rendom predation and the different proportions of

baits eaten by these birds are certainly not houogeneo us • The overall preferences for brown range ,in ordsr of decreasing

magnitude, from Vs obvious preoccupation with this colour

(though not as pronounced as in 2.10e to the sore catholic responses of C and D. The male blackbird 0 also ate a relative

excess of browns, but this effect was probcbly 1 not significant 1.69) • E, on the other hand, consurod relativsly iore

greens than opocted Cx 1 ) 7.33 P 4 0.001) 1 . 4Qr3OV5r, as mentioned above, all 6 of these baits were taken on the last two

1Besring in and the probable existence of heterogeneity within the overall results. - 168 -

days of the eper'iEent; during the preceding ten days, the bird

ate 144 greens cend no browns • This blackbird was wriiqus in the

hist097 of all ty '9 s I spaced-out' experinants it was the only

individual, that appeezd to acquiro E proference for geons.

Exporirent 2 .3a: cc arisen With 2.2

paGneity ketweendays. In contrast with their behaviour in the earlier experiment,, C and D did not exhibit strong brown

preferences at the start of 2.3a. On this occasion C showed no tendency to eat successively higher percentages of greens with

tima • Neither did any other bird, whether for greens or browns. The daily results noverthaleso appear to be heterogeneous, at.

least for C and D (e.g. note the relatively high proportion s of browns taken by C on 27.1.68 and by D on 24.1.68 9 25.1.68 and 2701 968 - Tthlo 43).

Overall responses, Direct comparison of the proportions ef baits taken by each bird in 2,3a with the proportions taken in 2.2

reveals that the overall behaviour of two of the birds had apparently changed sigeificantly (C, 42.9$, P (O.00lg E, 112(1) 12.219

P 4 0.001; D, X2(1) O.896 .S.; 69 (1) 1.58 9 1.S.). As can be deduced from Table 37, 8010 ction by C and E in 2.3a was now closer to random sampling. WO have previously noted that those two predators undetwent marked changes of behaviour during 2.2 • Thus brown preference decreased during the course of the ezperient, wh.1t E started to feed on browns only at the end of the Study, We should therefore Compare the predation by C in 2 .3a with that during each of his brown-orientated and gx'een.orientatod phases in 2.2 • The - 169 -

Tablo 39 Copaviuon botcon overall proportions oaten £d imprint - pocked (in bz'ckots) by blachblr& Ct,DEDG and somSthrmh T in ouperimants 24 and 2.3a.

BLACKBIRDS THRUSH C D E G 'F G B G B C B G B GB

2.2 97 113 140 57 158 6 24 5 7 117 (18e) (151) (30) (5) (32) (L) (1) (2) (6) (1)

3.944 3.256 0.0312 47.53 1) 40.05 net sign. not sign. <0.001

2,3a 101 21 178 59 72 15 153 16 (58) (19) (10) (6) (3) (1) (7) (4)

2 Z(1) 1.635 1.250 7.56 not sign. 40,01 P. not sip. - 170 -

first pair of proportions differ significantly (X2(1) 94.29 9 .

P < 0.001) whilst the second pair do not (X2(1) 0.095). SimilarlyD the proportions taken over E's first ton days in 2.2

differ significantly from the proportions eaten during the whole

of 2.3a (X(1) 26.560 P <0.001). The proportions consumed over the lastto days do not so differ. (X2(1) 1.68). Despite the heterogeneity between days, the findings imply

that two birds, D and C, tended to select approximately the same Proportions of baits during exuimont 2.3a as they did in 2.2.

Although a comparison of the overall results of C and E did not reveal such an effect, their respondse during 2.3a appeared to duplicate their behaviour in the later stages of 2.2 • The two

rsponsoa. of any given bird in the experiments tiers therefore somewhat similar. Yet the experiments were separated by a period of ton months.

(iii) int.pQckirg The grasping and dropping of a bait has been termed 'Imprint- 149 ). It is clear from the data given in Table 37 (page 164) that C was more prone to iaprintpeck than any of his opanions • This is a reflection of the bird's hesitant attitude towards the baits; each bait was usually imprint-pecked at least once before it was finally consumed or neglected. Table 39 compares the overell proportions of imprint-pecks with the overall proportions of baits actually eaten for those cases where the data are sufficient. In four out of seven

Instances the proportions do not differ significantly, suggesting that on these occasions the responses of the birds when irint- pecking were similar to when they were actually consuming the baits. - 171 -

Three birds (blackbirds C. E and song-thrush T) isprint-pecked in different proportions from when baits were eaten. In the case of

7 relatively too many greens wore imprint-pecked. This bird had a strong preference for browns but the data given in Table 42 a page 195, suggest that be was starting to search for greens towards the and of experiment 2.2. The data are few, but imply that the green itrint.podcing tended to occur prior and during this period of transition. The relevant data for blackbird G in 2,3& are also defiiont and in this instance relatively too many browns wore iLtt'-packed, all within the first half of the oxperimsnt. During

2.2, except towards the end, this bird had preferred greens and it is possible that the relatively high incidence of imprint-pecking in 2,3a was related to this fact.

The results given in Tables £12 and £13 (pages 195 and 196) suggest that some birds exhibited considerable variation between days in the proportions. of imprint..pecke. A closer examination of the results for Blackbird D in experiment 2.2 (Table 42) reveals that of the 30 green imprint-pecks, a total of 13 occurred on the second and third days, corresponding with the noticeable increase in the proportions of greens eaten by the bird.

The heterogeneity of the data for some of the birds renders the use of the chi-squared test in Table 39 statistically 'illegal'. For C however the problem can be partly overcome by partitioning the data into the first ton and last six days. If this is done (Table

38 0 p. 167) we find that during the first period the proportion of imprint-pecks differs significantly from the proportion eaten

14.039 P c 0.001) 0 while for the second period the proportions do not differ significantly (X1) 0.529). During the first period, - 172 -

as noted previously, the bird appeared to be gradually losing its preference for browns.

The behaviour of C. G, T and D therefore somswhat resembles

that of A and T in experiment 2.1 described in the lest Chapter.

That ia E the Start of large-scale predation on an unfamiliar colour tends to be preceded and accompanied by an increase in the incidence of imprint-pecks on this colour.

(iv) art-tors preferences

The data are not readily amenable for testing whether the predators were hunting by searching image, for the birds took few baits per visit in both experimante. 0 as may be deduced from the daily results given in Tables 42 and e3. Often only one bait was taken • Hence the data for each visit cannot be analysed for the existence of runs (if a given bird is hunting by searching image we would expect a tendency for greens to follow greens and browns to follow browns).

As already noted, howevero some of the birds appeared to take relatively large proportions of browns on certain days • Conversely, relatively high frequencies of greens seemed to be taken on other days. A superficial examination of the data revealed that the baits appeared to be taken in runs during these days. For example if a single brown was eaten by a bird on one visit, the chances were that a brown would be taken first during tho next visit.

The numbers of baits taken daily are not large enough for aningfu1 analysis, but we can utilise the whole of the data for each experiment. For each bird we can count the number of times a bait is followed by one of the same colour and the number of occasions different colours are taken in succession. Knowing the grand totals of baits taken it is a simple matter to estimate the - 173 -

Tthlo4O Eoiccitc 2.20 m4 2OO 0000d acid octod (in be1oo) numbom of tonao of Uko boft ooz foUwlng Smon md bn folloulne bo)

nd unliho bGlitQ ( cm io11o31j brctm cd bcm 2o11ot1nrj

SEQUENCES BIRD EXVI LIKE U1K0IXE X( 1 ) p

C 2.2 70 40 (77.35) (39.64) 0.0086 not

56° 21 (56.82) (20.18) 0°00t51 not elanif.'

2.3 85 20 (70.50) (33.50) 1.2934 not 815MM

D 2.2 117 65 (107.06) (74.94) 2.241 not Q1gnf.

2.3 134 67 (138.12) (02.88) 4.068 40.05

E 2.2 157 7 152.41) (11.54) 1.9243 not oi*if.

2.3 56 19 (53.62) (21.38) 0.3705 not

C 2.2 144 36 +2.3 (149.00) (30.96) 0.9935 not slanif.

T 2.3 17 (101.91) (12.00) 2. 3904 not - 174 -

expected msults for the above two catogoriess

if P r frequency of browns eaten

and Q " greens ' (where P+Q

then the expected frequency of the sequences bb and gg is (PP + QQ) and the expected frequency of the sequences bg and gb is 2PQ (where

b and g represent one eaten brown and one eaten green). Table eO glvns the observed and expected results for each

bird in the two experitients • The data for C have again been

divided into his first 10 end last 8 days of predation. Though only 1 of the 9 sets of data reveals signfioant deviation in the direction predicted by searching L!!agos, 7 of

the 9 othibit deviations in the direction ezpoted. Furthermore, the results taken as a whole indicate a significant tendancy

for the colours to be taken in 'runs' ((X)2 0 10,31, P < 0.01),

implying that the birds were indeed tending to hunt by searching image

3. EXPERXNT 2.3 (CO1TD.). PARTS to d AND at TRIORPHXC POPULATIONS

This short study wo a prelude to those to be described in

the next Chapter. Its a.ts was to determine the responses of birds to a third morph (khaki), whon they were already familiar with greens and browns • Since the new type of bait was intermediate

In colour between the standard greens and browns, the results should provide a clue as to whether birds can discriminate botwoon such similar prey. - 175 -

Qronolm The dates (168) of presentation tere as follows:

Expt. 2.3 (coat.)

a. From 4 February to 7 February: 90G + 20K + 90B

d. From 8 February to 16 February: 67G + 67K + 67B

a. On 17 February : 90G + BOX + 203

In other words, the khakis were first introduced at relatively low 'frequency and this was increased during d and a.

1atorisnd thode The material for the new baits was made by thoroughly mixing equal amounts (by weight) of standard green and standard brown dough. After a uniform khaki colour had been attained the bits were manufactured in the s=3 iiiennor as described in Chapter III. These baits were also used in onperimente described in Chapter IX. Plate 4 (page i) shows the various shades of khaki that were used in these later experiments; the intermediate fore (abbreviated to K In the Tables of the present Chapter) is labblled as 14'. All the birds involved in 2.3a and b also preyed on the trimo2r'phic populations, although E was absent during the duration of c. A new male blackbird (I) from an adjacent tarritory started to feed on the baits during the preeenta'on of the 1:1:1 populationog in common with other intruders he was liable to attack from C and D.

(a) Results

(i) Sutmary of data

The ecp1eto daily record for the numbers of baits eaten and iint'poekod is givn in Table 43 pago 196 • Table 41 aummarisee Table 41 xperiment 2,3 (cont.) c,d,e. Trimorphic populations (K--khaki) : grand totals of baits eaten and imprint pocked by blackbirds (c,D,E and I), robins, dunnocks and house sparrows, Expected predation, on the basis of the proportions offered is also shown.

mci. B L A CK B I 1?. D S )Dn. dates C D E G I ROBINS DUNNOCKS H. S. (1968) G K B G K B G K B G K B C- •K •B G •K B C- K B C- K B .3c. )O0 4.2 13 0 54 17 1 57 - - - 37 7 11 - - - 15 0 16 31 1 7 - - - + -7.2 9) - (si) (i) ------- ------OK + 9:2:9 30.2 6.7 30.2 33.8 7.5 33.8 - - - 24.6 5.5 24.8 - - - 10.4 2.3 10.4 17.6 3.9 17.6 - -: - )OB - -

57G 6,2 1 1 88 33 25 30 5 6 7 31 29 29 2 15 13 23 7 29 48 6 29 12 0 3 + -16.2 (i) (2) (95) (2) (2) (i) ------- ------37K + 1:1:1 30,0 30.0 30.( 29,3 29.3 29.3 6.0 6.0 6.0 29.7 29,7 29,7 1060 10.0 10.0 19.8 19,8 19.8 27,7 27.7 27.7 5.0 5,0 5.0 57B

)OG 17.2 14 0 0 94 0331 6 8 1 9 7 2 6 2 113 1 1200 + '5) - (2) - (i) ------- - - - _•, - - - -

) OK + 9:9:2 6.3 6.3 1.1 5.9 5.9 1,3 3,2 3.2 0.7 6.8 6.8 1,5 8.1 8.1 1.8 4.1 4.1 0.9 7.8 7.8 1.5 0.9 0.9 0.2 OB

a - 177 -

those data and shae the overall results for imprint-pecking and predation by the five blackbirds and the robins dunnocks and house

8parrOws.

Pm&ticn

The new morph clearly tended to be disregarded when first

presente4. Only the outsider, G. took khakis in any number. The

other two blackbirds (for sorts reason E was absent during the whole

of c.) concentrated on greens and browns, in particular the latter (favourite) variety. Robins and dunnocke also neglected khakis. The lost species appeared to prefer greens this belief is

strengthened by the results for 4, and a,

Men the morphs were presented in equal frequencies blackbird D started to take. khakia, and overall it took approximately equal proportions of the three rnrphs, though on a daily bats the results appear to be heterogeneous • G and E and the newcomer I tended to prey

randorly. On the other hand, the three species of •smell' birds and blackbird C continued to neglect the khakie • The lost bird was now concentrating almost entirely on browns. Although Lower baits wore taken, the trend observed in d. seems to be continued in a • That is, blackbirds D. E, C and I tended to predate in a random manner, whilst blackbird C. the robins, dunnocks and house sparrows tended to underpredate the khakie, which were now presented at a higher frequency than before.

. DISCUSSION

This is an appropriate point to consider the behaviour of the two I Turdus species when feeding on green and brown baits • The - 178- dincoion toill &co bo cmaorned tiith the moults described In to pvioue Ctor as ofl a thoo of the paont no.

(a) PooaLb2,e ceot of bzn foznGz.

All tioo bido (AO B ond T) in experimontsexperiments 2,la cad 2.1b (ch.vn) and 2oz' bi'do (CO D9 T and C) in oiporint 2.2 (thin Ch.) took move bone than opootod on the basis of rcmdom prodat1m. One bird (E) took relatively too iany gmens in the latter oerie3nt but in 2.2a this birds as tcU ao the other qmdatova invold (CO D and 0), preforred bvmo., In other tor min out of eawin birds, had profbmmew For browns and the oonth profored btzo on one occasion and Smens on another. This result reflects thoeo Siven in Chapter VI tbe elewin eporim3nts save data for p3dst5.on by aroupa of blackbird and in ewiry case brone tore taken more often than opeoto4 assming no selection. The remainder of oVerimnt 2.3 provides further ovidenca that blaebir&i pfer brown to green. Mma a third 9 khakL o, morph ae proaontod in 2 • 34 t'o of the blackbirds (C and D) took rXatLvely mom ome, whilst a third (G) took relatively toro greona than expected. Khakis were airost oonpletely neglected by C and Pan4 tiere slightly overpz'odated by 0. Mien the three colours wore offered In equal qtenoieo in 2.34 C was the only bird out of fio that eboted a distinct ymferonce for browns predation by the renaming binds was pore ou, laee Chapter IX will also present evidenan that blackbirda prefer buom to green. In the oernonte described there, sin out of

RMI - 179 -

seven birds tended to take the browner varieties from a mixture of

baits coloured 'arious shades of green and brown • The seventh bird however showed a vary definite predilection for certain types of green baits (see p. 22. From the above we may conclude that most (though not all) blackbirds and aongthrwshes encountered pzofomf brown. This

behaviour tended to persist within individuals (for four wonths t, in the case of Tj for eleven months, for C, D and G). It was recorded in both Edinburgh (Scotland) and Reading (England) separated by a distance of aom tOO miles. In view of the consistency of these findings we should enquire into the causes of the colour preferences • There are five main possibilities.

(1) Relative cnsiouousneas

The expericenta described in the last and the present Chapter were carried out on grace and it is conceivable that browns were prefoz'ret because they were relatively core conspicuous • To the human eye browns appear to stand out core than greens, but not to such a degree as would be predicted by the results for oongthvush T, for exailo. In addition, the other species (robins, dunnocke and houso sparrows) in opericsnts 2.1, 2.2 and 2.3 had distinct preferences for green (see Tables IV and VIII, Allen 1967, for 2.1 and 2,26 Table 43 9 this thesis, for 20). If browns had appeared conpicuou to all birds then we would expect the lest three species to also have chosen browns. Finally, similar ezperizsnte carried out on soil and snow backgrounds also revealed strong brown profbrenew in blackbirds and eengthrusboe (Ch. VI, p. ll8 • Tahing all these facto into consideration we cay conclude that th9 two species in question did not preferentially select browns becausq they - 180 - were more conspicuous than greens.

Taste

The results could of course be explained on the .basis of differences in taste. but the evidence does not support this view.

No great differences were detected in the hidensity experiments described in Chapter IV: if anything, greens were prof z'rc3d. Furtherore sortie birds in the present experiments took browns to excess without ever oanpling any greens • The most obvious cases concern blackbird A and songthrush T. Clearly these birds were not neglecting greens on the basis of taste; they did not take any greens at all during the-grater part of 2.1.

Inflate preferences .

If the colour preferences were due to genetic causes then the situation would prasuxaably be an example of behaviour palymozphiam or 'polyathism' (Chance and Russell 1959) • Various workers (e.g. Kear 1964, ffeidmann 1965) have studied the colour preferences of now-born chicks of various birds but the individual responses appear to be relatively constant within species. It seems that the innate pecking responses of Turdus species have yet to be studied.

Persanent acguirodpreferoncos

Rabinoiitth (1968) and others have shown that birds can acquire colour preferences through.experience early in life (see P-36, this thesis)• These preferences are rigid and peroanent. It is well known that parent blackbirds food their young with earthworms (Snow 1958) and is conceivable that the brown preferences are so acçid, Green preferences could be explained on the basis of early diets - 181 -

containing 8 preponderance of green proy, for examplo b catorpiflare.

(v) TetOrax acuie4 aferences The blackbirds and thrushes may have already been using

searching images for brown prey prior to my experiments • Earthworms

(Lübrious app. and A. lolob.ophora app.) wore occasionally taken during the studies, sometimes in preference to baits • This was

especially apparent at the start of experiment 2.2 and accounts for the low predation by the two birds involved. Perhaps significantly,

they both took a majority of browns on these days.

Heppner (1965) has Shawn that Turdus. migratorius hunts for

the tips of earthworms in their burrows mainly on the basis of visual cues, probably shape or colour. If T.merula and T.Philomalas

also hunt earthworms in such a manner then they may well tend to attack bra= objects within an area whore they normally hunt earthworms • It should also be noted that the shape of a bait is not so far removed from that of the tip of an earthworm. Such a

searching image may be 'area-associated' (Cz'ose 1967, 1970) 9 that is it may be restricted to sites whore worms are found s In other areas the birds may switch' to other searching images. For instance green searching images may be used when hunting in a

tree which from past expericrce is associated with green caterpillar's. In the absence of appropriate behaviour experiments it is

difficult to evaluate the cusea of the preferences. Suffice it to say that in practice all five possibilities may be involved,

(b) Sorcin maeQ

te must now examine the evidence that the birds were hunting by searching image. In particular, were the strong brown preferences -182-- of T and A in the second pert (b) of oxporimnt 2.1 caused by the

acquisition of saarchin .4 images during the first pert (e)? If this were the case then the preferences in 2.1b nit be expected

to have been stronr than the rasponsen of naive birds to eiti1ar

9Gz lB populations • Four such 'naive' birds wors sncontered these vem blackbirds C c D, E and G of experiment 2.2a. Thsh 7 An this oxperineni had already experienced baits in 2.1 and should

strictly be neglected from the cozparicon, as should the Tudus

Individuals in 2.3 since these had all participated in 2.2.

In fact two of the four naive blackbirds in 2.28 took a clear aaority of brotns on the first day of predetion and these praferances were persistent. t5o have already mentioned that this behaviour mñy have been related to searching for earthwores, The

preferences of these birds wore not so strong as those of A and 7

in 2.3.b • During the six days of 2.1b g A took 1 green and 130 browns and 7 took 0 greens and 176 browns • For the equivalent

period at the start of 2,2;ic and 1) took 17 Smena g 75 browns and 56 gmens q 31 browns respectively. This suggests that the preferences of A and T had been enhanced by b wn.pecific searching

But in viety of variability in behaviour, we must be wary

jf drawing a firm conclusion from these results • It is possible

that A and 7 possessed strong brown preferences for reasons

unconnected with conditioning to brown baits. Additional support for searching itmege is that the birds were

dstecting the prey in Irmal when they tore foeding on 9U

populations (the data from trimorphic populations were too eccty for aaeningful analysis). Those results suggest that the birds tended to search for one colour at a time and wer capable of switching frem using one searching isga to using motheva that io, - 183 - they 'eitthod attention' (cg. Dwkins 1971b).. Similar .reolto have been obtained by L'ou (1064) for chickadees hunting dyed stmflower soods and by Dawkins, (1971b) for chicks feeding on orange and green grains of rice.

Cwoze (1967 0 1970) on the other hands found no evidence for

'rane' in cwo preying on populations of red, black, and yellow

17nsoel ehelle in oqal proportions. He o1ggosts that this inability to 9brm searching ineges lowers the officiency of the predator, for the crews took longer to find each prey in po1yco,hic populetiono then in equivalent Eonoxrphic populattone. This cannot be refuted by my work, sines no nonotaorpbic populations wore presented and tiie eceurerQonts were not takenaccurately; never holeos the birds appeared to be wnting by searching, Image 1, for they were taking the baits in 'rune'.

Cre3e was his data as support for 'adaptive po1yxophis&

(Dobehenoky 1951) which izp1ies that polymorphic populations are 'ore fit than those which are conotoz'pbic • However the theory of adaptive polycozDhiom has yet to be eubstantiateds a po1yhicie taintainod by selection can occur in a population without affecting its adaptation (Cain and Sheppard 199Th) • Cae 'e suggsztion is nevertheless interesting and demands further investigation. But if colour po1yorphisin does increase protection from predation the question as to bow it is maintained still z'et3atns unanswered. Xf pro &ttore uno unable to fore eoarching images than epostatic aelection cannot be izp1ictod.

Cc) Relevance of iuintckin 'He noted that the onset of largo.i..scato predation on a new colour is often associated with an increase in imprint-pecking On - 184 -

Ws colour. This was particularly noticeable for birds A and 7 in 2.1 but the birds of 2.2 also behaved in such a manner. It may follow that the strength of this typo of imprint.'.pecking is directly proportional to the otrength of the preference for the original colour, but the data are not conclusive. L. Tinbargen (1960). and Nook e.t aJ.. (1960) contended that a prey is not finally accepted until it has been subjected to frequent thanca encounters, The increase in green imprint.pecka prior to the taking of greens supports this view of delayed acceptance, but indicates that the prey need not necessarily be eaten duxing this period. If the animal has a hard integument it may well survive imprintpecking. Similar findings were made by Beukema (1968) when he introduced Dro$opbila larvae to sticklebacks, accustomed to feeding on fe woz.. Results e!dn to mine were also reported by Harliles (1933) in exporimonto designed to reveal the colour preferences of wild bix'ds notably tits (re Maj,jr L. and 2e4eu L.) and a nuthatch (Sitt guz'qea L.) • Coloured peanuts were presented in various proportions. Unfamiliarly coloured peanuts were pecked but not eaten at firot and were taken only after repeated pecking. Assuming that greens taste the same as browns, we must consider why' they were often dropped after being picked up.

Presumably the fact that they were the 'urong colour' overrode any positive (or neutral) taste stimulus. A comparison can be drawn with the reactions of predators to Batesian mimics.

Brower (1958 9 1960) has shn that an avian predator familiar with a distasteful model may sototimee accidentally pick up a mimic, but release it again. In this situation the predator belatedly - 185 -

'z.Ueec' that the prey is the 'wrong colour' even though it is not distasteful. But the ',xong colour' is the fmiliar one. Such findings toro obtained for Scrub Jays (CyjgCitta 4aerulescens)

feeding on mimetic butterflies (Brotior 1958) and for starlings (Sturus ultez'i) feeding on mealworm 'models' and 'mimics'

(Ber 1960) • In Batesian mimicry a rare palatable prey species is protected by resembling a co=on distasteful prey species, in non-mimetic co],wr polymorphism a rare palatable morph may gain protection by virtue of it being different from a cozon variety.

In each case an accidsntal attack on the rare type need not necessarily recult in its death. Imprint-pecking was not limited to the occasions mentioned above; both colours tended to be imprint-pocked (at random?) at other times in the eper'iments • Sometimes the cause was r3cogPisebles

for eemp10 a oudden frit at the moment when the bird was picking bait, but in most cases the cause was not Obvious.

(8) Divreity of behaviour and its ipplicatons

(1) Divorsit. of behaviour

The birds clearly differed in their ovo rail responses to the baits • The strengths of their preferences ranged from T's almost total prooccupation with browns to the more random predation of E. Other forms of behaviour also revealed idiosyncrasies, for example, with regard to the imprint-pecking (e.g, the behaviour of birds prior to the acceptance of novel baits, -p,157f. and p.170f the ecesoivo Imprint-packing by C, p.170 ; the imusual form of imprt-pocking ozhibitod by A and T. P.158 ) • Indeed my long periods of observation soon convinced me that each bird had a !character of its own. The recognition of these idiosyncrasies was partly based on subjective Impressions. Thus 0 was by far the least afraid of ean while B was the most timid.

Other differences were potentially measurable but had to be

noglectod because of the difficulties of attempting to observe

many things at one time (see p. 60 ). One example is the speed at

which the baits were eaten • The territorial lotmerst of the plots (blackbirds A and B in 2.1; C and D in 2.2 and 2.3) tended to chase out intruding blackbirds ond eongthrushes, and these interlopers thus tried to grab as many baits as quickly as possible. Blackbirds

G. I and to a lesser extent, E. were such birds. It may be significant

(although the sample is small) that these birds tended to be the least selective; with such a premium on the baits, their preferences were relinquished. The 'resident' birds on the other hand had more time to exercise their choice • The behaviour of eongthrush T does not fit the above argument because this bird was almost invariably attacked when in the plot and yet was the most selective • Do intra- and inter-specific competition have different effects on prey selection?

The heterogeneity of behaviour was probably caused by a combination of both genetic and environmental factors. The overall proportions taken by a given bird were no doubt influenced by the strength of its colour-preference. In other words the causes of the diversity of response may be identical to the causes of the colour preferences which were discussed in (a). It is also likely that a bird's first (chance?) encoustero with the baits influenced its subsequent selection (of. Tinbergen 1960). Uo have seen that this certainly

1This bird had the audacity to beg for (toast-coloured )crumbs at the breakfast table! - 187 - seems to have been the case for A and T in experiment 2.1a. These birds apparently acquired searching images for brown baits in addition to their natural brown preferences. Variation between the responses of individual predators has been reported by other workers, for example with regard to unpalatable prey. Thus Cott (1940, p.286) writes: "individual toads appear to differ widely in intelligence and rate of learning" in response to warningly coloured bees. Comparable results have been obtained for palatable prey and were discussed in Chapter II. Smith (1967) with sticklebacks and Pough (1964) with chickadees, both produced similar results for predators feeding on artificial prey. More recently Mrs. Pat Miller (thesis in preparation) has detected great

variability between individual quails feeding on artificial prey. We should also not overlook L. Tinbergen's (1960) original discovery

of the diversity in the responses of individual tits hunting for

insects in pine forests. On the other hand, the three rudd used in the experiments of

Popham (1941 9 1942) appeared to show almost identical responses to corixid bugs (Elton and Greenwood 1970). However this result is not necessarily contradictory to those mentioned above. The three fish might have been taken from a population showing great

behaviour variability and their apparent uniformity could have been a result of chance errors in sampling. A repetition of Popham's experiments would be a worthwhile investigation.

(ii) Apostatic selection We have seen that some Turdus individuals took browns alone,

whether they were presented with 9G:1B or 1G:9B populations. Such behaviour is of course not predicted by the hypothesis of apostatic selection. When selection by groups of birds was examined, however, - 188 - wo found that the cocbLxsed effect ta appacnt1y froquency

pendant, asoming ccetcnt co1ectio, in 13 out Of 14'931' pori!!eflt3 (aptor VI) • oa avo'aga individe1 birds tended to ht for the coner co1ore.

Tho3 may be tt3 min receono ftr the disparity between the findinep for individual birds (ChQpta3 VII end VIII) and groS of birds (Chapter VI) • Firet q birds sh as A end T o with very str brown proferncoe may have bean in the minority in eaøh of the 13 OmperlimentO with the zsult that their effect was ovorralad by the, frequency-dependant tendencies of other jividuals • Second, those birds which did start by taking the rora co10uge alone may have fowd this behaviour uaprfitab1e and cttchod to oearchina for the coion varioa • One factor which may aid this cheno is cociel facilitation. Various ioz'kers have shown that some animals can learn from the oMerienco of others and this has bean termsd 'obsorvational 1Larnin' (for a source of rsfaroncee to the psychological litoraturee sac John al al. 1968) • Birds have often boon Shown to learn from watching others during feeding; oap1es of species that have been investigated aMS Chloris gloo, Xlcpfoz 1961; various Fringidae s, I(ear, 192; PQeCCr oiodc and o1Lebs Turner 1965; dozestic chLcko Tolman 1968; Colui

Norton l971b) • These birds feed in denee flocks and it has been suggested that communal feeding and roasting improve the 2badincj prospects of the individual (ward 1965 s, Zchcvi 1911,

Murton 1971c) • In View of this evidence VC miit pdict that social predators htrnting for rsrs prey itotre may lose their searching images in preference for more profitable ones by matching and - 189 - copying the behaviour of others. Apparently just this has been recently demonstrated by Murton (1972b) for wcodpigeons Columba pabus feeding on seeds 1 .

There is no conc1uivo direct evidence that social facilitation contributed to the overall fraqusncy.dependent results of my

9:1 experit!snts. In experiment 2.Th for example,, blackbird B took greens whilst her mate A continusd searching for rare breins, oven though the birds were often in the plot together. In experiment 2.2, blackbird D took a majority of rare browns on the first day and on subssqizeñt days tended to concentrate on greens, while his mate blackbird C gradually changed to accepting greens throughout the psricd of study. There is no clear evidence that the alterations in the behaviour of either of these birds was related to social facilitation between them. An important consideration is that the blackbirds did not appear to be particularly 'social' feeders. Even when both resident birds (A and B in 2.1 C and D in 2.2 and 2.3) were in the plot they kept very Duch to themselves and often fed in different areas.

The fact remains that most of the 94 experiments gave indications of frequency-dependent selection for predation by wild passerines on tase. In some experiments (1, 3.1 9 3.3, 3.4 9 4) large numbers of birds of various species wore involved and it is possible that social facilitation helped to ensure that the birds hunted on average for the coioner colours • In each of the other experiments (3.2, 5.1, 5.2, 5.3, 5 14 9 5.5, 6.1, 6.2) less than six predators were involved,. sontiees only two (experiments 5.3 and 5.4),

1Iurton's (1971b) paper had not been read at time of writing. It is quoted in his (1971a) article. - 190 -

Yet the apostatiC effects of the predation in the last group of experiments seem to be no less marked than those of the first (Tables 19 and 20, pages 11, 115). The possibility that Turdüs individuals learn from others that certain prey are more profitable has yet to be substantiated. Social facilitation is potentially an Important tool for strengthening apostatic selection and some polymorphic species undoubtedly fall prey to birds feeding in flocks. But what of songthrushes and Cepaea, for example? The songthrush is usually an wISOcla]. bird in the British Isles and is unlikely to hunt Cepaea in dense flocks. In this situation it is improbable that songthrushes acquire profitable searching images by direct observation of other individuals. However, there remains one

other possibility. Although songthrushes may search for snails independently of one another, they may tend to break the shells at a common site, particularly if few anvils are available (Morris

1954). A thrush with an unprofitable searching image may be able to form a new one on the basis of the shell remnants at the anvil.

If this were true it would be an indirect form of social

facilitation. This hypothesis may not be so fanciful as it may seem; songthrushes are known to respond to empty snail shells

(Morris op. cit.) and it is the colour and pattern of the pieces

that would be important, not their irregular, non-snail-like

shape (p. 37).

(iii) Predators and area effects In (I) we remarked that the results may suggest that blackbirds

ae more selective within their territories. Since the data also

show that individual birds vary considerably in their preferences It - 191 -

follows that pairs of birds may also differ in their overall

responses. L.Tinbezgen (1960) in fact demonstrated that pairs of

tits in adjoining territories show different prfarences for insect species. In view of this the morph frequencies of a polymorphic

prey species might be markedly different in adjoining territories even when there are no apparent differences between the habitats. The degree of permanence of such areal differences would than depend

largely on the pomuanence of the territories • Thus since songthrush

pairs tend to keep the same territories for a year and sometimes

longer (Davies and Snow 1965) we may predict that discontinuities

of ph-frequency in gMga could last for a similar duration.

To my knowledge noone has attempted to find any relationship between morphfrequencies in 2gaea and territories of songthrushos.

Creze (1967) has extended this argument in an attempt to

provide an explanation for 'area affects' in Oapaea as first detected by Cain and Currey (1963) and later confirmed by other workers (e.g. Carter 1968 Arnold 1971 for Se2aea; Clarke 1968 for Partula taenjata), Area effects are characterised by areas of constant morph frequencies which may change within distances of 200 in or lace without reference to the environment. They may extend for 10 sq, km or more. Croes (1967) suggests that they may be the result of area-restricted predation by certain species with large hunting areas • Songthr'ushes are ruled out because of their' relatively small territories and Cr'oze suggests Corvidas as the most likely candidates • He admits that his is'gross speculation° and the long existence (from at least the Neolithle v Cain and Cuz'rey .: 1963) of area effects argues against this explanation.

A more likely hypothesis based on selective predation is that area effects are, associated with the distribution of a particular species of predator (Cain and Currey 1963, Croza 1967 s O'Doneld 1968). - 192 -

No such correlations have yet been detected, but if it is true that songthrushes, like blackbirds, prefer browns (and the evidence from Chapters VII and IX implies that they do) then

this species may be implicated. Some area-effects in Cepaea are characterised by high frequencies of brown morphs (Cain and

Currey, qp.clt.). Perhaps such localities are correlated with

the absence ot near-absence of songthrushes. Of course, at this stage we know little of the colour

preferences of songthrushes in nature. Is the preference for

brown widespread geographically? Even if it is we cannot

eliminate the possibility that yellow, for example, is preferred

to brown, for the types of colours used in my experiments were

strictly limited. In fact, the preferences of songthrushes might well vary

from place to place, perhaps depending on the colour and abundance of alternative prey (p. 181). This leads to another

possible explanation for area effects 1 . In places where earthworms are common thrushes might be expected to take relatively too many brown morphs if Cepaea is present. Where there is an - abundance of palatable yellow (or even green) caterpillars the birds may tend to overpredate the yellow morphs. Distasteful prey

might also affect morph frequencies in Cepaea. Thrushes which have experienced aposematic yellow-and-black striped prey may be

prone to avoid yellow effectively banded Cepea. It follows that area effects in Cpaea might be related to the

distribution of other prey species which contribute to the diets of its major predators, bearing in mind that the songthrush is not the

1 1 am grateful to Professor B.C.Clarke for drawing my attention to this hypothesis. - 193 - only predator of Cepaa. A1thogh Cain and Currey (1963) were unable to detect any obvious relationships between area effects and the distribution of other species the situation warrants a larger ecological survey than they were able to carry out.

(s) Zntzoductiài of third morph

The results of 2.3 c s, d, o support many of the findings from the dimorphic 9:1 etperimeuts • The birds were again shom to be diverse in their responses. Equally important., the khaki morph was found to have an overall advantage thon first introduced.

This suggests that the birds had become accustomed to searching for grecns and browns and overlooked the novel prey. If so, then

this would be further evidence for apostatic selection. However D the data do not exclude the possibility that khaki ties neglected because of some factor independent of searching images. Further eperirxnta with khaki morphs were doomed necessary.

. SW1ARY OF CONCLUSIONS FROt1 EXPERI2EN1'S IN CHAPTERS VII AND VIII

Data were obtained for the individual responses 0f blackbirds and a oongthrush. In summary, the main conclusions tore

Host of the birds had frequency-dependent preferences for

browns.

One blackbird selected in an spostatic manner,

Large-scale predation on an unfamiliar colour was preceded

by an increase in the frequency of 'imprint-pecking'.

The birds differed markedly in their responses to the baits

and in other aspects of behaviour.

There was evidence that the birds were hunting by searching

image. - 19'4 -

APMjDjjt C: TABLES 42 AND 'iS ftU moulto for ooLtnto 2,2 md 2.3, Maptax, VIII Table '#2

Experiment 2.2. Full daily results for blackbirds C, D, E. G and songthrush T. Imprint-pecked baits are shown U) En in parentheses. The numbers of visits (v) per day are also given. '.4

180 greens + 20 browns B L J C K B i II D S THRUSH Date C D E G T (1967) G B Iv B v G B v G B v G B-

20.3 ( 4) + 3 ( 2) + 9 8 1 (i)+ 7 5 - 21.3 (.2) + 1 ( 3) + 18 17 (s)+ 14 (i)+ 4 12 - 22.3 ( 4) + 1 ( 14) *15 19 (8)+ 15 5 14 - 23.3 ( 9) + 4 ( 20) * 16 20 (i)+ 9 ()+ 5 13 1 0 1 - - - - - 24.3 ( 9) + 2 ( 23) + 11 17 (l)+ 7 7 10 ()+ 4 05 - - - - - 25.3 ( 16) + 6 ( 6) + 6 11 ( 1 )+ 10 3 9 ()+i2 .08 - - - 0 8 26.3 ( 6) + 2 .( 28) + 9 11 ()t 11 3 11 (2)+24 0 8 - - - 0 (l)+14 2703 ( 18) + 3 ( 17 + 8 11 (i)+ 14 4 ii (3)+1s 0 10 - - -(2)+0 16 2803 ( 22) + 5 ( 16 + 8 .15 ( 2) + ii (i) + 3 14 ( 7 + 22 (i)+o 11 - - - 0 17 2903 (22)+14 6 * 3 14 () + 10 2 10 (3+17 0 7 ..0 ()+o 1 0 17 30.3 (32)+16 3 + 2 16 (i)+ 6 8 12 (5+16 0 9 - - - • 0 11 31.3 ( l8)-,.15 11) + 2 12 (2)+ 14 (l)+ 3 12 (4)+13 0 6 - - - (3)+2 4 1.4 ( 11) + 16 ( 2) + 3 17 (.2)-i- 10 2 9 17 0 7 0 1 1 W+3 16 2.4 ( 7) + 3 0 5 - - - 5 1 2 .(l)+4 1 3 -. - 3.4 4) + 6 3 10 8 1 5 (i)+ 9 5 6 20 3 8 2 14 Grad totals 97 113 1202 140 57 147 158 6 80 24 5 13 7 117 (184) (151) (30) () . (32)

* One of these baits was imprint-pecked in an abnormal manner (see p.158)e GT)14.'I 06 + L 6 I 9 °LT 0 0 •t I I T 6 i 9 L L I c_ 'c 1. o (T 6 o () . ()

3cL. DU ID 69E 3 L _ i L1 ci 6 __ Ti L (c oc _' _) _() _ c () T cc__ I()i(t_ ) 7 b c o 0 c-i: i ' LG 9 ; , L (I) (T) t r z't) o 0 ( 91 L •t I 0 61 6, 1 6 L I j7 9 9 9 . L 9 (1) 6 I L CT 0 • 1. . ii (t) o o V ST ji I L I _ 9 - e c c ( -c) 9 01 . 91 (ci) 0 0 tI 0 9 (IC 0 L E:I L I i — .- - - - (I) 0 0 3 9 9 ( -t) I id t L 6 9 0 1 I 9 t 9 L iiL(LI)O 0 STPTL9+ 9 9 . L . (i) oJqL9+GueGJL9°p -ti: •t? 0 9 T I -— — — - . LI (ot) o o T1 L_i -ri: fi10i - - — fl_L1 .. . CvL _i 1T(r - rco _ c-( sT2opIr 0 t OT ( L - - L i ol vt e - r• 91 ';i (rc) o c () — - — — I 0 ii g c: o _ - — IL ... . c;-t . tE9t (L).o () 9 . — - — - Li o_t 01 0 - .— c 0 • I L i7t8T 0 ILI(t7E)O () To+ - °j7 IT c •t7,O•i — - - — - -. - — - — — 9 01 0 c:9 ()o c A E )I OA I NJ A : i 0 -A [ >1 0 A )I A a >1 A a x c A N 0 .

s -poq. p _ zj —:- —:------::-- 69 -- 90 [ ' 9' ' L ' _(' _0 (c ) ff 8 0 . L L 9 0 TI i(E) 0 T CL i o L 91 OT 01 91 • 0 6 8(6) 0 () T I'1r 0 _tT1ZXD 0 0 o 9 ° çi Ot -E () _ ZL1) L91 6 (°) _ tjc: 6L T (Gt ICI — — .- 0 01 1 (i) 01 T i (i) IT ;-• -r- 6 • 0 9 1 c[ ( ) 0 6 0 6 0 : c o ci . 0 L T c c-t oi (9 ) (t ) ioc . s 6 z CT (1 6:c(9) 9 T°6a Cc o c -i: 0 1 c o c • • i " - .- - - - - c I ct I 0 1 01 1 CT (t) 9() 11(9) I ° 8? . c o I .- — 1 0 1 - . I 9 I 9 g . () iL . T 9 6 c: IT 91 ci(c) .r9 - -. - .— — .- — - •t I i/l F, 0 •t c-i: 6 L Go 2T (9 - - — - - — - — - (i) 9 T 0E 9 L I IL (L) - L 0 (i) 12 () ' (i)• -t . ; 0 IT (T •tiE 9 171 6 t 6 (6) t - - - - — —- - — - (i:' 01 9 z L ET (1:) 01 T?? - - — -. - — .- - - , 0 ' (ci): 9 0 IT (F- F j- - - — ( -1:) . 11 (I) L 9 : - - — TO '. — - -- - - — - - - . — - - -- •:- L 3 (T) 8 161 — — — . - — - - - - - — - -1:-i: i' ( -c) 91 - i.e-i: .— L •- •------— -. _t () — — - c-i: -1: 61 ( - .- .- - — - - - - - — - — - - — i71 9T(T'L 19T _ A _ A A _[ f) A _[ • ) A A 0 (96T) . My" JS (7)I Cl eAT OT &t je J c;t ;o oi ez D soU2u3-1 0 9 4ICO I UJOX put' (,ji) c :'ooump ' (TmpTAtpuT su . qo pT ' ' ' '3 spxqYoq. IO --ITP _ g I q x 3 V 'I . . ------197 -

CHAPTER 19 FURTHER ZXM-RIMIT-S WI

IMEWRIATE .COLOUR S - 198 -

1. INTRODUCTIOI?

The results of the lest experiment of the previous Chapter hosed that a third, khaki, prey can be overlooked by predators.

In this Chapter we examine further' etudies involving baits of

various shades of Ithaki. The work wad, divided into two parts • In the first, an

atteipt was made to determine the specificity of searching images. The experiments of Chapter V clearly showed that gild birds can

discriminate between standard greens and browns after appropriate training. But as pointed out at the time,. polymoztphs of natural prey are often much more visually similar than the standard dimorphs used in my experiments. In the work to be described, this

discrepancy was rectified by giving birds the opportunity of

discriminating between baits of relatively greater visual

similarity than standard greens and standard browns. If colour polyaorphisms are in fact maintained by apostati

selection then how did they originally evolve? The second set of experiments examines the possibility that predators are capebla not only of maintaining, but also of promoting polymorphisms in their

prey by ercising dieruptis selection. Consider' the hypothetical evolution of polymorphism within a 'species' of bait that exhibits continuous variation in colour ranging from brown to green • This

variation is distributed normally around a 09aa colour composed of brown and green in equal proportions. Ideally we should present such populations to wild birds • Obsarvaticns of the response of - qq -

:• "" '1 .. •.' ':-: (... Ile

7 . : •-•..t - ..- • 4'. ,'.. --.'-...-.- "'7.. ve _.• ..• p . •• •

-. -; .' :

. I.

4 • i-; . tc

Flat. I;

Appearance* of the nme morph. as used in .xpsriMsnts described in Chapter IX. Increasing gr.n*issfroa left to right. The bait. are each 0.7 an long. - 200 - the population to selection should detomiine whether discontinuous variation is evolving. This could occur if, for instance, the predators concentrate on searching for the commonest colours.

Such behaviour would imply the existence of searching images of a highly specific nature.

Two experiments were carried out along the above lines • In practice it was nocaseaxy to limit the number of types of baits. Nine varieties were offered and it may therefore be argued that the populations were already polymorphic. Even so, 1 often had difficulty in distinguishing between the different shades of khaki.

The two groups of experiments are first described eepaz'atsly.

They are then collated in a general discussion and are compared with other researches particularly those of Cross (1967, 1970) • The khaki colour-types were basic to all the experiments and are dealt with below.

2. PREY

The nine varieties of bait are shown in Plate 4. Table e4 gives details of their optical properties. Each morph is numbered according to its difference from standard brown in arbitrary units • Thus 0 represents the base colour; k, the intermediate khaki as used in experiment 2.3. Chapter VIII; and 8, standard green.

The baits were made by mixing standard green and standard brown dough (see page 55) in the appropriate proportions by weight. For example. ,1 1 9 contained equal quantities of green and brown dough, whilst 7 ,s were mode from seven parts of green to one of brown. Large numbers of baits wove made at the same time and their' method of manufacture was as described in Chapter III • The non-chromatic - 201 -

Tablo 44

Colows of the nine gmen t khaki and bxom belts used in onperimnts described in Chapter IX. The data were determined by use of the Eun5oU Book of Color (1966). I an grsteful to Professor Bryan Clarke of the Genetics Dopartttent, University of I1ottinghan for supplying these data.

Colour Hue Valu2/cbroma of bait

o 5 YR 6/6 (otandd brotrn) 1 7.5 YR 7/6

10 YR 6/4

3 2.5 YR 6/4

5 y 7/4

10 Y 6/4

6 2.5 GY 6/4 2. 5 07 6/6 8 7.5 07 7/10 (standad green)

Note: The hue of a colour indicates its relation to red, yellow, green, blue and purples the value its relative lightness or grey value and the chroma its strength. - 202 -

properties of the baits were kept as constant as possible and sore no different from those of the standard greens* and broms used in previous experiments.

3. EXPERXENTS 1 /4ND 2& TMINIG

Xtro4uctoxy remarks

Those experiments were similar in design to thcse described

in Chapter V; birds were first trained on one colour and then presented with 1:1 populations. The most Important- difference as that the baits employed in each experiment wore visually

move similar.

P1c

bur eperimonte were carried out and eli involved initial training sessions with 4 1s. They were divided into two separate series which wore planned as follows:.

Serjeo 1 (July 1969) comprised two enperinsnts 4t. separate sites. After being trained on k's, the birds were presented with 1:1 populations • In one oporimont those populations contained t' and 8's (standard greens). At this site the birds were next trained on B's and then offered 1:1 populations as before • The second exporiient was of similar desLSm,, except that O'

(standard browns) were used in place of S's.

Series 2 (August-September 1969) also involved two experiments at different locations • They differed from those of series 1 in that the baits wore visually core similar. In addition to one eoritnt i corporatod the presentation of 2 1 0 1, whilst the - 203 - Table 45

Sites of specificity training experiments, giving O.S. grid references.

All experiments were carried out on short grass.

Expt. Site

1.1 . Lawn in front of Zoology Department, Kings Buildings, Edinburgh' N.S. 268707

1.2 Golf course near the Poultry Research Centre, Kings, Buildings, N.S. 263705

2.1 Lawn in front of Chemistry Department, Kings Buildings. N.S. 266707

2,2 Playing. fields, Liberton Dams. N.T. 272704

Table 46

Dates of training and presentation of 1:1 populations in experiments of series 1 and 2. All dates are inclusive.

TRAINING ON 's RE VERSAITRAINING

Expt, Colours Dates of Dates of . Dates of Dates of training presentation training presentation (1969) of 1:1 pops. (1969) of l:1pops.

1,1 & 8 5.7 - 137 - 17.7- 24.7 - 12.7 15.7 23.7 25.7

1.2 & 0 5.7 - 13.7 - 17.7 - 24.7 - 12.7 16,7 23,7 25.7

2.1 4 & 6 18,8 - 25.8 - . 28.8 - 409 - 24.8 27.8 3.9 5.9

2.2 4 & 2 18.8 25.8 - 28.8 - . 4•9 24.8 27,8 . 3.9 5.9 - 201.1 - other Gt710fl3d 6 0. Thtte oaoh ouporl=nt con3ist3d of ttio partst a. fVaining with Ilb e fc11ctod by raoeettion of 11 op1at1cno; b.. oL..tining nd preontation of li populations.

Cc) aeri1e 4 thod (I) ia All four opoz'i3nt8 took place on ohort grass at differont oitoe tLthin i.ilo radius of the Department of Zool*Mro University .- of EdinburCh (sea Teblo 45). Each =Vrimental plot tivae a square

coictinj of ttiony$ivo tuo.

edze

Table 46 Sivao the dztos of the poaiode of familiarization

and presentation of 1 I populations. The training s essions all

looted for one week ai4 eippoiinately 2000 baits we= cons4

at each site. Poplationc centa.tning 25 of each of the relevant

colamm wero peontod dewing the fizt two or tbe days follodnrj tho ondb of the training 000aiono. They wave observed

QOntLnOO17 CMd tco maintainod in the zsiol mmnarv, with p.tocetont of eaten baits • The times of poeentatian OE 1l opa1ations oz given with th rze1to in Tthloo 48a and

48b.

(iii) PM.4zAqM Blackbirds pcyed on the baits in all four oztorittrnto.

In addition house oezowe and 4unooko tor inrolvod in ozorints 1.1 amd 2.1 and the pzdetion effected by these

'ceall' birds two calculated in the usual fashion (see p • 85). Ecticatoc of the nwbors of birds involved ozo given in Table 47. - 205 - ,

Grand totals of baits taken from 1:1 populations after

training, by blackbirds (Bi) alone or with house sparrows

(Hs) and dunnocks (D). The estimated numbers of each

species involved are given in parentheses.

Expt. 1.2 Expt. 2.2 Expt. 2.1 Expt. 1.1

Bl (3) BL (4) Bi () Bi (5) + + HS (6) HS (6) + + D (2) Ii (2)

.2 i. .L a. AFTER TRAINING ON 4's

Grand totals 80 61 101 78 96 77 128 59

X()(deviation from 1:1) 2.560 2.688 2.087 25.460 P N.S. N. S.. N.S. 0.001

b, AFTER REVERSAL-TRAINING

Grand totals 45 62 52 68 59 77 64 74

X(l\(d1atb0n from 1:1) 2,701 2.133 2.382 0.725 / p N.S. N.S. N.S. N.S.

x ( 1) (heteroene1tYCD 5.245 4,927 4.466 16.001 between grand totals)

P <0.05 <0.05 <0.05 <0,001 (d) Results Tables 48a and 48b (pages 209 and 210) give the results for daily predation on populations presented in series 1 and 2 respectively. Analysis by chi-squared shows that the proportions taken per day do not differ significantly within parts a. and b. of each experiment. These data have therefore been pooled and are given in Table 47. This Table does not present the experiments in chronological order (from left to right) but in a sequence according to the relative greentteSs of the baits that were offered with the 4 1 s. Furthermore, the results for predation by blackbirds and 'small' birds in experiments l.la, l.lb and 21a, 2.1b have also been amalgamated. This seems a legitimate procedure because there is no significant heterogeneity in the overall proportions taken by the two groups of birds within the halves of the relevant experiments (see Tables 48a and 48b). There can be no doubt as to the effects of training. Direct comparison between the two grand totals for each experiment reveals that the proportions differ significantly in every case (bottom row, Table 47). Moreover, all eight grand totals deviate from

1:1 in the directions expected on the basis of conditioning. The results for experiment Lie, however, are the only data that differ significantly from the expected ratio.

Further evidence for conditioning can be obtained by examinng the data in greater detail. The graphs shown in Fig. 11 illustrate the relationships between the percentages of 4's taken over every five visits. Since the numbers of predated baits were small these percentages are subject to sampling error. In spite of this, the graphs have striking similarities. - 207 -

Figure 11

100 I .Q)

•1 50

0 I'..., I. I 10.1 24.7 ' 25.7

B Expt2.2(4&2)

50-

I I - 25.8 ' 26.8 ' 27.8 49 5.9

100 C Expt2.1(4&6)

50. \,...... •

I 0 25.8 26.8 27. 8 9 59

100 0Expt1.1(6&)

50 •N

0 13.7 ' 14.7 ' 15.7 24.7 ' 25.7

DATE

Percentage of L's taken over every 5 visits from 1:1 populations after training in experiments 1.2 (blackbirds), 2.2 (blackbirds), 2.1 (blackbirds, house sparrows, dunnocks), 1.1 (blackbirds, house sparrows, dunnocks). Besides the experiments involved 0's, 2's, 6's and 8's respectively. - .208 -

The highest percentages of 4's were generally taken immediately after the first training periods. This is especially true for predation by blackbirds in 1.2a, 2.2 and 1.1. The blackbird and small bird results for 2.1 also fit this generalization, though the effects are not as pronounced. These results are mirrored by those fox' the second parts of the experimentsi. All the birds began by taking relatively low proportions of 4's; that is, they continued to search for the familiar colours. This may imply that the tendenôy to take the conditioned colour declined with tims (as was noted in experiments described in Chapter V) but in fact none of the data exhibited significant decreases in the proportions taken over ovary five visits (Speax'man's rank correlation test).

4. EXPERIMENTS 3 AND 4; 'N0RIALLY-DISTRIBUTED' POPULATIONS

(a) Introductory remarks

The training experiments demonstrated that birds can distinguish between baits of similar colour and these findings stimulated the inception of the two investigations described below.

The first experiment (experiment 31) was essentially a short pilot study designed to. test the method. Experiment 41 was a more prolonged investigation.

As will be seen, the principle aim of the work was to detect changes in the structures of the populations resulting from selection by predators en moase • Nevertheless it was possible, especially in the second experiment, to record the behaviour of individual Turdidao.

11be experiments are numbered in sequence with the two series of training experiments. - 209 -

Tab1o. Specificity trainiag-experimntes series l Daily results for psdation by blackbirds (Dl) cind small birds (S) j, 10o. house sparrows and dtnnocks..

periment 1.1 ps'imant.l

Data Times of Predation Dates Times of Predation observations BI S observations Dl (B.s.T.) (B.S,T.)

. 4 9 8 8 40 - - -

a • AFTER TRAINING ON 4's

13.7 1405s.1730 34 8 24 10 14.7 1240.1645 31 17 14.7 1900-2040 22 7 12 11 15,7 13111430 26 23

1597 1925-2050 .2 13 13 10 16.1 1300-1642 2 3 21 Grand totals 79 28 49 31 Grand totals 80 61 (hat. 3O5 2.263 0 (bet.beteen 1.831 bet,aan day (Nose) days) (M.S.)

hot 3.355 41)(between - ,T,'s)(N,S,)

b e AFTER REVERSAL TRAINING

24.7 . 1930-2140 26 30, 10 14 24.7 . 1020-1215 8 20 25.7 1951-2130 19 20 9 10 it 1410-1645 16 21 25.7 0950-1415 21 21 Grand totals k5 50 19 24 Grand totals 45 62

(hat. 0.048 0.138 (hat .between 3.198 2 'between days)(N.S.) (N.S.) 4 trials) (H.S.)

(hat. 0.121 4 between G.T.'s) (N.S.)

- 210 -

Tth] 48b

Specificity training experiments z series 2. Daily results for predation by blackbirds (Bi) and small birds (5) i.e • house sparrows and dunnocks.

Extormnt 2.1 .EezIment.. 2.2.

Date Times of Predation Date Times of Predation observations Dl S observation Bi (B.S.T..) (B.S.T.) 46 4 6. .. 4. 2 - - - - -. -

AFTER TRAINING ON o'o

25.8 1930-200 21 11 12 7 25.8 1059-1715 37 18 26.8 1850-2106 23 19 19 16 26.8 1021-1521 31 30 27.8 1920',2146 14 15 7 9 27.8 1000-1530 33 30 Grand totals 58 45 38 32 Grand totals 101 78

1.930 1.318 )(bt.S between 3,830 ) (het.between 2 4 days) (N.S.)(14..S.) 4 days)

(hot. between 0.069 4 G.T.$) (H.S.)

AFTER REVERSAL TRAINING

4.9 1850-2131 19 29 7 16 4.9 1100-1630 24 37 5.9 1708-2101 21 20 12 12 5.9 1005-1650 28 31 Grand totals 40 49 19 28 Grand totals 52 68

41) (bot.bettteen 1.209 1.867 41) (het. between 0.804 days) (N.S.) (N.S.) days)

(bet.between 0.256 G,T,'s) (N.S.) - 211 -

W .Basic dosip of. experiments

The second experiment differed from the first in that it was preceded by a period of training with i's.. The reason for this will become clear later, In other respects, however, the two studies were essentially smiler in design.

Prg

Nine morphs were used. These ranged in colour from 0

(standard bom) to 8 (standard green) and are illustrated in

Plate e, page igg.

Presentation of p4pulations

Birds were first offered populations with the morphs in frequencies that approximated to the normal. distribution (see

Table 49) . The frequencies of the two extreme colours were correlated with the tails of the distribution, whilst 's constituted the mean (and mods).

Table 49 Initial composition of 'norci&.ly'.dis tribute d' populations expressed in percentages.

orph 0 1 2 3 4 5 6 7 8 % in population 2 6 11 20 22 20 11 6 2

The positions of thebaite within each metre-square were determined in the usual manner (see Chapter XII) and were marked onto a copy of the grid. Playing-cards wore then used to ensure that the colours of the baits wore randomly distributed within each quarter of the plots of the first and subsequent populations. - 212 -

A mark on each card indicated the. colour it represented. The

relative proportions of the marked cards and their total number

reflected the constitution and number of baits that were to be distributed. After being thoroughly shuffled the cards were dealt out and the ensuing order of the colours was systematically

superimposed on the positions of the baits marked on the copy

The populations were continuously observed during presentation

and were iaintained in the normal manner e by frequent replacement of eaten baits.

Changing the couositions of, the populations

All previous experiments involved the presentation of populatMa

with constant proportions of two or three colours • In each of

the present studies the compositions of the populations were periodically altered. This was carried out after a certain fixed

percentage of baits had been removed by the birds • As a consequence

of selection (or random sampling variation) the relative numbers of surviving baits were likely to differ from the original proportions.

A new population (or 'generation') of the original size (see below) was then constituted,, such that it contained the morphs in the

same proportions as had survived predation.

The point at which the composition of a given population should be changed was governed by the following considerations. Ideally, the absolute number taken should be as large as is practicable.

This would reduce possible errors due to sampling variation (caused, .,fox example, by differences in the numbers of visits by different birds) • Since it was hoped to change the population about once a

dcr, the main limiting factor was the rate of predation. Thus the populations were changed when 100 and 70 baits had been taken from - 213 - experiments 1 and 2 respectively. The sizes of the populations were respectively 1,000 and 500. The presentation of so many baits would, of course, be impractical under the standard conditions of two baits per square metre. Instead, the smaller 'presented-populations' 1 were considered as samples of the larger populations. In addition to the 100 baits offered in experiment 3, there were 900 in reserve, whilst in experiment 4 9 200 baitswere offered with 300 in reserve. For authenticity, the percentage of baits taken from the population should be of the same order of magnitude as in predation on natural populaiions. The proportions used, 10% (expt. 3) and 14% (expt .4), are not unreasonable values, in the light of our present knowledge (for general examples, see Lack 1954). Few data are available for predation on polymorphic prey, but Cain and Sheppard (1954a) estimated that thrushes eliminated about 8.6% of 10,000 individuals in about a fortnight. After the specified number had been taken, the baits were removed from the plot and the new frequencies were estimated.

This calculation often resulted in the computation of fractions of baits. For the purposes of estimating the constitutions of the presented-populations, the fractions were rounded-off to the nearest whole number.

(c) Practicability of method The setting up of the presented-populations was a slow process initially, simply because of the variety of baits involved. After practice the reactions of the human 'predator' improved. It is also interesting to note that at first I was unable to identify

In the following pages 'populations' refers to the overall numbers of baits (i.e. 1,000 and 500). The term 'presented-population' is self-explanatory. Teble 50 Eoitont a. Mmmbcra o baits in ttho1e populationo (N n 1000) dto' S Ganomtiono of coleotion, TO Tcblo io Occh Population 2 ra1ueo for the deviation of Ito ecWpoolticn fran the original distti its =m t (with 95% confidence Units) end Ito otonderd deviation a,

22TEBLTXODI 0- 1.- .2

0 20.00 20.94 21.36 21.68

1 60100 57.30 48.89 47 096

2 110.00 104.73 98.93 8948

3 200.00 199156 191.19 179.81

4 220.00 218.26 222.44 220.54

5 200.00 203.97 215.33 228,10. a 6 110100 117.90 123.96 129.22

7 50100 55 1 20 S4.14 57.05

8 20.00 22.04 23.68 25.95

4$) (devition from 1.67 7.88 17.57 • generation 0)

95% confidence limit 4.107 4.136 4.205 4.274 (upper)

4,.000 4.029 4.099 4.168

95% confLdenco limit 3.893 3.922 3,993 4.062 (lower)

0 1.733 1,730 1,715 1.719 - 215

accurately the colour of a given bait lying in position on the grass, unless it happened to be one of the two etremee the

familiar 0's and 8's. Identification of the colours was necessary whenever the baits were removed frOm the grids. On picking up a bait I attempted to assess its colour before tz noferring it to its

corresponding Compartment in a tz'5y • It was at this stage that errors of identification became obvLot. With time I appeared to become more proficient but never attained 100% accuracy. Unfortunately

It was impracticable to keep a record of my responses.

(d) Prej ftinar, otu4yt exez'irnt 3

atsa1end..nthode. The experiment was carried out on .a lawn in Portobello,,

Edinburgh, in September 1969. The size of the lawn limited the presented-population to a 10 m x 10 in grid containing 100 baits. As already mentioned, the size of the whole population was 1,000 and its composition was altered after 100 (10%). baits had been removed. This figure was adhered to as closely as poseib1e 4 so

as to reduce error. The first, nox'na11y..distributed,popu1aticu (0) was presented on 18th September. By the end of a week, three further populations

(generations) had been presented. Predation was restricted to blackbirds alone and there were at least two males and two females.

Individual-.birds could not be reliably recoiised.

Reau.to and,ftscupsion.. Response-of po,u1ation :tó selection

The compositions of the four generations are given in Table 50. Comparison of each population with the original composition shows - 216 fka

, uormsi diotbuion' oporiisit S. of popu1L1iofl ooe (boko lino) nd ciftoz (wbokon lino) 3 ConoraZions of 9o1Octio.

240

220

200

Im Ice z - 140 C 0.

0 120 0 ci

C 100

0 E. 80

I 2. 1 .4. Colour - 217 -

increased deviations tith tiie.. After the third and last

generation Of oe1eCtion this departure was statistically significant (see valuos of x2 given in Table 50). Figure 12 illustrates graphically the relativo frequencies

at the beginning (generation 0) and and (generation 3) of the

experiment. Tho shape of the latter graph gives the impression

that it is skewed towards the' 'groner 9 half of the distribution. Its modal colour is now $ and each of the green colours (5 7.9 B) is at a hie frequency than it respective 'brown counterpart

Assuming that the populations have continuous 'distributions, we can calculate t.héfr means and standard deviations. These are given in Table 50 with the 95% confidence limits for the moans. Each successive generation has an increased mean, supporting the view that the populations were becoming 'greener'. Populations

19 2 and 3 have lower standard deviatiOns than that of 0 > implying loss variation about the means.

Earlier experiments (see Chapters V e V1 9 VII, VIII) revealed that blackbirds and thrushes generally prefer 's to Va when presented in "epacod-out' populations.. The psent results are undoubtedly partly due to a tendency for the birds to take the brownar colour. However, the situation is complicated by at least to other factors. Firet, there is some evidence for frequency-dependent selection. That is,both of the rarest colours, 0's and B's, apparently enjoyed relative protection from predation and consequently increased in abundance. In the absence of frequency-dependent selection and in the presence of preferences 'for brown, we might expect the Ots to - 218 - Table 51

Pooled data for predation by female and male blackbirds on populations (0,1,2) of experiment 3. Two pairs of birds were probably involved. -

Females

Colour Numbers taken of total

0 3 1,7 1 .27 15.7 2 35 20.3 3 52 30.2 4 38 22.1 5 10 5.8

6 5 . 2.9

7 . 2 1.2 8 0 0.0

172 99,9

Males

Colour Numbers taken of total

0 2 .. 1,7

1 1 . 0,8 2 12 10.1

3 . 24 20,2 4 26 21,8

5 . 25 21.0 6 11 9.2 7 17 14.3 8 1 0.8

119 99.9 have decreased in frequency, as did the l's, 2's and 3's.

Now due to the close visible similarity between the 8's

and 7 1 s, we might predict that the latter also would have proliferated within the population, particularly as the other two types of green (6 1 s and 5 1 s) exhibited such a trend. In fact the 7's suffered a slight negative change. This brings

us to the second factor complicating the results given in

Figure 12: the behaviour of the predators. The result for the 7's is probably associated with the heterogeneity of the birds'

responses, as we shall see.

Behaviour of the birds

The three presented-populations (0 9 1 and 2) 9 being one-

tenth of the size of the overall populations 9 did not differ significantly in composition. It is therefore legitimate to pool the results of predation on these populations. Accurate data for individual predators cannot be obtained however, since the blackbirds were unmarked. But the birds could be classified

according to sex and a comparison of the results for males and females gives an indication of differences in behaviour.

Table 51 gives the pooled results for predation by the two sexes and these data are also presented as percentages.

The distributions of the two sets of data differ significantly

(P c 0,001, Kolmogorov-Smirnov test). Predation by the females is heavily biased towards the browner morphs. Each

of the green types (5, 6, 7, 8) was taken at a lower frequency

than its brown counterpart (3, 2, 1, 0). Amalgamation of the - 220 -

Figure 13

Site of plot in 'normal distribution' experiment 4 (March-April 1970), showing boundaries of blackbird territories. Compare with Figures lOa and lOb, page 147.

\ I ___

-V I

I - ----I M 1 f

KEY outline of plot. 0 —bush 0—tree f —flower bed wn long grass -territoria/ boundary x-observation point

Sca1 /e 0 5 /0 4feb-es - 221 -

four types in each category shoes that 71 greens were rauovod

as opposed to 156 bruvns • If a cinilar calculation in carried out for the males, we find that 516 graea and 39 browns were

taken; in other words,, greens vera preferred. Inspection of the data reveals that an excessive prodationof Z'° probably contributes to this ra3u1t. Plaid observations eeeed to Indicate that a single bird was concentrating on searching for .1'o.

() Eperiint 4 The pilot experiment demonstrated the feasibility of working with nine typos of similarly coloured baits and paved the way for a mora prolonged study with individually recognisable birds.

(i) Oaterialsand izothcda This second experiment took place at my Reading home during

1arch April 1070. The reader is referred to P-148 for details of the environment. Though of similar concept,, this study differed from the previous one in the following points:

A. Prior to the presentation of generation O the birds vera familiarized with ,'a for one flook • It was hoped that this procedure would nullify the birds' overall bias towards broina • All the participants of the later part of the experiment were observed feeding on the training diet.

B • An attempt was made to quicken the response by lowering the population from 1000 to 503 tthilst raising the selection to 20%. In fact a value of 14s was ultinatoly used because the rate of predation was slow. - -. -. -- -1. - - - I V S S C4A A 5 c;Lj S CA Experiment., 2 values for its goodness-of-fit to the original distribution, its mean (with 95% confidence limits) and its 4. s'andard deviation a

GENERATION 0 1 2 3 4 5 6 7 8 9 10 • 0 10,00 11.76 13.10 12.73 13.74 13.53 15.12 13.05 15.21 15677 16.2 1 30,00 35.28 38.18 35.76 38.61 . 36.42 . 40,24 41.35 38.8o 36.84 3.S •2 55,00 56.47 .56.20 54,13 54.13 58.92 64.94 69,16 73,60 68.74 69.1 3 100100 100.00 98.02 95.21 98.47 92.74 91.18 89,41 87.90 84.61 81. 4 . 110.00 102.35 89.50 82.00 73.41 77.69 73.07 66.97 68.73 67.70 60.E 5 . 100.00 91.76 96.63 102.80 102.30 102.00 95.10 97.52 102.80 104. ° 97,34 6 . 55.00 55.29 57.14 59.81 61.34 59.16 53.16 55.34 54,01 52.53 55. 7 30.00 35.28 38.18 . 43.78 4.5.10 45.06 51.84 49.18 43.24 46.83 50,4 8 . 10.00 11.76 13.10 . 13.88 12.82 14.78 15.38 18.05 21.04 27.91 26,

X ( 8) (deviation from generation o) . - 3.73 10.47 17.56 25.26 23.11 40.18 45.53 46.69 55.94 76. 95% confidence limit (upper) 4.107 4.138 4.165 4.260 4.235 4.250 4.'206 - 4.221 4.188 4,273 4, 4.000 3.979 4.001 4.095 4.068 4.081 4.032 4.047 4.013 4.097 . 4.1 95% confidence limit (lower) 3.893 3.820 3.837 3.930 3.901 3,912 3.858 3.873 3.838 3.921 . a . 1.733 1.818 1.875 1.894 1.920 1.925 1.989 2,000 2.009 2.030 2,C

GENERATION (cont.) . l 12 i3 14 15 • 16 ' 17 18 19 20 o 18.6.3. 19.85 19.23 19.41 21.20 23,03 21.03 23.24 . 20.14 19.68 • 1 . 36,73 41.45 41.61 44099 49.53 47.30 43.46 36.51 33.31 . 29.36 2 66.83 65.43 74.69 12.59 72.11 73.03 75,77 80,95 84,05 77.78 86l6 82.54 82.79 89.45 84.63 85.02 80.41 61.97 52.53 49,34 4 57.31 55.27 52.82 42,85 47.07 47.90 46.49 48.14 56.23 58.21 5 99.45 96,40 92,92 81.28 74.01 63,80 60.36 67.72 61.01 0 61.59 6 52.83 48.73 43.09 44.37 38.44 38,06 37,36 36.39 33.16 37.31 7 53.35 56.52 57.55 17.67 72.21 76.56 84.55 87.66 96.56 102.80 8 28.71 31.88 35.26 37.41 40.80 45.33 50.49 57.42 62.40 64.27 X( 8)(deviation from generation o) 91.93 118.39 144.93 192.07 234.94 285.62 351.06 432.84 520.48 561.33 95% confidence limit (upper) 4.336 4.341 4.311 4.34.9 4.339 4.378 4.499 4.626 4.744 4.867 - • 4.151 4.152 4.118 4.151 4,136. 4.171 4.289 4.414 4.532 4.657 95% confidence limit (lower) 3,966 3.962 3.925 3,953 3.933 3.964 4.079 • 4.202 4,320 4.447 or 2,110 2.163 2.195 2.261 2.317 2.363 2.389 2.420 2.424 2.406 - 223 -

The size of the presented-population was 200. Seven Turdidas were involved. All the five blackbirds

were colour-ringed or otherwise recognisablo. Two iiringed

aongthrushee also participated in the otperiiuent but it was

impossible to distingUish between them. With the exception of the

male blackbird E 1 4 none of the birds had been involved in previous ci) expsricemte at the site Frequent territorial displays occurred

thtoujiout the experiment and revealed the approxLeate arrangement / of the territorial boundaries shown in Figure 13.. Robins also fad on the baits occasionally but since the numbers taken were small, this species was disregarded for the purposes of the experiment.

Population 0 was presented on 21st March in a 10 m x 10 in grid

() sited as shown in Figure 13. By 15th April 20 generations of

selection had been completed ,.

insults discussion

apcnse.of Pc 02n to , s2jjctL on The complete data for the compositions of the populatians are

given in Table 52 9 p.222 • As indicated by the values of X 2 0 the distribution rapidly becama significantly different from the original

frequencies (bearing in mind that P m 0.05 when 28.13, for 8 degrees of freedom) • The Table also gives the meens and standard deviations of the populations. These data show that the distributions

were gradually becoming greener, but with more and more variation C about the moans.

Ile muat examine those changes in more detail. The evolution of

the populations can best be visualised from a graphical presentation

'For the previous history of E t see Chaptert!.

- 224 -

Fi gure 14

'Normal distribution' experiment 4. Morph frequencies in populations for every fifth generation (0, 5, 10, 15, 20).

110

105 / / '4 100 S I 95 I /1% I/I 90 / II \ II 85

80 75 'I 0 0 it) 70 11 65 20 o60 0 \ *r, C=. 0 CL \ ' 50 C (\\ \ .. 45 \•\ '4 40 \ .15 \ \\ 35 '4 \ 30 S '4 '4 \\ '4 *10 :" .

15 -y / ' .5 •/ 10 1 \ 5- 0 2. 2. I Colour - 225 - of the data in Table 52,. This is done in Fig. lL • In order to avoid confusion in discerning the main trends 0 every fifth generation is shcn. Certain features are immediately obvious.

The first relates to the increase in variance noted aboves the centres of the distribution become progressively 'eaten-away' during the course of the otperimsnt. For the first twelve visits or so, the range of this depression is confined mainly to the i's; in contrast, the two most similar colours, S's and S's suffer appreciably less change. Apparently, then, the birds were tending to concentrate on eating t&"e. On the basis of previous 'training' experiments (particularly those described in the first part of this Chapter) we can presume that this effect was a result of the birds' previous experiences with this colour. By the and of the experiment, 3Is and 'a were apparently also being taken in relatively greater nuriers 0 for their frequencies decreased vithin the population.

Correspondingly, the '7's and 8's underwent two- and five-fold increases in numbers As in the previous expørimnnt, such effects could be caused by an overall aversion for the greener colours. If this were the case then we would expect similar results for the

6's (a colour that I myself often confused with 7 1s). In fact, the 6's actually decreased in numbers • We will presently see that this effect, was probably the result of the idiosyncrasies of a single male blackbird.

Behaviour 0f the birds

The birds each took too few baits for meaningful conclusions to be derived for predation on single populations. However, there - 226 -

are no significant differences in the proportionswithin every

three (or two1 ) generations and the results for individua12 predation over these periods can consequently be amalgamated.. Table 5e (pages 256 and 257) gives these data in full and Figures

15 to 21 (pages 229 to 243) present them graphically. In the Figures the predation on a given colour is expressed as a percentage of the total number of baits eaten by the relevant

bird (or birds., in the case of the threshes). Specific types of lines are used for each of the predators and the average

compositions of the populations over every three generations are shown by the boldest lines. Predation by the female blackbird M is omitted from the first two Figures since totals of only five baits and one bait were taken.

Fig, 15 for predation over the first three generations (0 - 2)9

shows that all the depicted birds tended to take more Os than expected, In other words,, all the five points az's above the mean percentages of 4's offered. Furthermore, male blackbirds 3 and L

took relatively more 's than any other colour whilst the thrushes

appear to have been concentrating on both 4 1 3 and 5• The most striking result is that of 3 which took more than twice as many 41a as its next most favoured colour, 5. The male blackbird E and

the female K took relatively more 3 9 9 than e'a. Over generations 3 to 5 (Fig.16) the birds still tended to

take more 4 I than expected 9 though again this was not necessarily

the preferred colour (K took relatively more 3's, for example).

1Since 20 populations were presented they can be divided into six groups of three and one of two (generations 18 +19), 2Pz'edation by the individual oongthrushos could not be recorded because the birds were indistinguishable. By necessity the data for the thrushes refer to their combined efforts, and in the following pages the two birds ax's treated as ono this is simply for ease of presentation. Tab 10 93

baito tzoB by osw ovovy th (or to) SaBomtsaw In Ewrimat 29 with oa& calcamod gtnpo data r_ivm In Tciblo 54 • Thc avoraga imoons &md atemdard Qrmzu iof tho pc atQ..po4atio wo alco

B I R D S

E J x Too O 2 3.711 0.373 4.093 ' 0.266 3.407 0.482 3.710 0.590 3.833 0.468 3.903 0.295

3 9 t75 O.GIO 4.774 + 0.310 4.037 ; 0.528 4.000, 0.622 4.135 0.54 4.091 0.267

6 - U 3.598 0.540 5,321 0.374 3.862 0.636 3.71 0.759 2.929 0.808 3.269 0.472 4.020 0.280

0 - 11 2.933 0.523 5.275 -' T 0.260 3.600 0.661 3.778 0.664 3.737 * 1.306 3,939 0.436 t.142 0.290

= .14 3.783 s 0.720 5.192 0.365 3.76 • 0.698 4.484 0.635 3.727 0,567 3.105 ; 0.575 4.140 0.39 -4.

15 17 3,593 0.892 4.700 0.907 3.208 ; 0.809 3.256 0.797 8.031 T 0.649 4.578 0.581 4.091 T 0.040

10 0 370O s 1.102 9.036 0.702 3,294 0.719 3.500 1.040 2.583 0.883 2.870 0.811 4.320 0,351 - 228 -

In contrast to the other eight colours,, e's were underpredated by

only one bird. Surprisingly, peraps 9 in the light of the preceding • three generations, this predator was blackbird J • As can be seen from Figure 16, J was now concentrating on searchiflg for 's and

these baits were taken over three times more frequently than any

other. The differences between the proportions taken over

generations 0 2 and 3 - 5 are statistically significant (P .4 0.001. - Xologorov..Smirnov test on data shown in Table 5l) • In other words

there appears to have been a distinct change in preference by this

bird. The five remaining Figures show that J was persistent in its

treatment of s. These were always the most preferred baits,

though 'a (Figs. 17 0 18 and 19) and I's (Fig.17) were sometimes also taken at relatively high frequencies. It is obvious from the graphs that this bird preferred the greener baits V the browner ones and this point is well made by Table 53 which shows

the mean colours taken by each of the birds.

With regard to the remaining birds, the most consistent trend from generations 6 - 9 (Figure 17) onwards, is a tendency for the graphs to be skewed in the direction predicted on the basis of preferences for brown • This is especially noticeable in Figs. 17

and 18, whilst reference to Table 53 shows that in contrast tp the

means of the average prssented..populations • the birds tended to

take samples with means below 4.

5. DISCUSSION

(a) Stimulus gqneralisatioq and the specificity of conditioning The training experiments demonstrated that wild birds can

(cont. p.244) - 229 -

lee 230 to 243 : Figuros 15 - 21

Pex"contago3 of the eight types of baits taken by each bird in 'normal distribution' experiment 4 over avery 3 generations. The percentage average compositions of the populations are also shown. - 230 -

y to F ig. 15 .

predation by blackbird E

p , J

- I, it It K

ft U it L

X )C H 9f songthDush3B

average composition of population

62 -231 - Figure 15 60 - 'Normal distribution' experiment 14 58 Generations 0-2: percentages of the eight types of baits taken 56 by each bird and percentage I average composition of 54 population.

52 I 50 I\. 48

40 1 • I. 38 . . C A .2 36 1' I ' 'I• I , A 'A 4 I o CL 32 o S A C 30 I + / •i 28. ' A • 1' I o 4 CL \ \ A E 26 Uo - •', , 24 . %

22 > . A1 \ C rd I

20 A 0•' ro /4 18 IIJ o /41 16 CL :1 I r 41 p

14 AgI .41 / C 12 t .

rd I I 10 44 I • + 8 •/ \ \ - .4 ' C I ) 6 a) / 4 2

Q ! 2 3 4 5 6. 1. Colour - 232 -

Key to Fig.16

predation by blackbird E

- 7, 7, 7t J

-- it 77 " K

7t 7? L

it it x x )C K X songthrushes

average composition of population

62 -233-

60 Figure 16

58 'Normal distribution' experiment 4. Generatinns 3-5: percentages of the eight types of baits taken by each 56 bird and percentage average composition of population.

4446 I\ 42 / \ 40 / 38 g36 I\ 75 34 I ft 32 1-o • 3O I I o E28 (I) ' o I CL 26 o K 24 '

I )' 22 I - 20 CM / I 40, 18 C 16 0 1/ \\• \ -D 14 Li :" C (Ti 4 12 . I 4 K

K\jL K K - 10 K 4' O8 (0 /'!

U 6 /

.___ .- '•'' 4 . 4*t I 2 - N 0'• 'I. 1 Colour - 234 -

Key to Fig. 17

predation by blackbird E

It J

it It It K

It It It

it It II rvI

if if songthrushes average composition of population -235- 62 LJLgure 17 60 'Normal distribution' experiment 4. Generations 6-8: 58 percentages of the eight types of baits taken by each and percentage average composition of population. 56 bird

0C 36 ro

CL 32

28 ;. o \ -26 /

24 / :\ \ •,/ 22 > ' q 20 •'! )yr?Syt

i \\ L \\\

8 / / , + - — • / /1 / \ 4. j 6 ' I / . S \• \

: : 0 2. 1 2 3 ± Colour - 236 -

Key to Fig. 18

predation by blackbird E

It It U J

- U It If K

It It L

It I? It - . M

It songthrushes

average composition of population -237- 62 Figure 18 60 Normal distribution' experiment 4. Generations 9 - 11: 58 percentages of the eight types of baits taken by each bird and percentage average composition of population. 56

38 C 36 I \\• 5 34 CL a32

30 .

: A Io I ° 24 I a , ' I •-- - -. I ' .'•i .' 22 a, .1> : I•

16 it

14 'I

\\40 __ / .•./ 2 .kI

2 ! 2. Z. Colour - 238 -

Key to Fig. 19

predation by blackbird E

11 U U J

U it it 1<

U U I I L

U I I t

tv ' songthrushes

average composition of population - 239 - 62 Figure19 60 'Normal distribution' experiment 1+. Generations 12-14: 58 percentages of the eight types of baits taken by each bird and percentage average composition of population. 56

42 40

_5 34 CL H CL 32 15 30

22 \\

20 ,;/i:•;;;:." ;/ /

I Ilk 16 CL / 14 -4 ...... 12

:f•2' /

2 1 . ± Colour - 20 -

Key to Fig. 20

predation by blackbird E

If U J K it it

U L

U U U M

U U songthrushes

average composition of population -241- 62 Figure 20 60 'Normal distribution' experiment 4. Generations 15-17: 8 percentages of the eight types of baits taken by each bird and percentage average composition of population. 56

34 £1 :: 0 cL. 32

0 30

E / 24

20 /

'. C : 18

16 CL

c 12

rd J A V / 4,X\ \ 8

K ) / N

2 ! Colour - 242 -

Key to Fig. 21

predation by blackbird E

U it Vt J

Vt it ?t K

-- U it Vt L

Vt M

it it songthrushes

average composition of population - 243 - 62 Figure 21 60 -- 'Normal distribution' experiment 4. Generations 18-19: 58 percentages of the eight types of-baits taken by each bird and percentage average composition of population. 56

42 - 38 C 0 36 (Ii 3 ci 34 ci.0 32 0. 30 0 S. I U) 28 0 0 E 26 0 U CU 24

I- a,> 22

a, 20

S. a, U_ L (U 16 0 c 14 (0 12 a, -x (0 -a 10 a, (0 S. Ca-, U L_ a-(U

Colour - 21iZ - discriminate between two poy of oimi1ez colour. Sieila' eou1ta wero obtained for ooe of the birds in operinnt though in thic case the aopho wozi even ieee alike • tSo must now ow!aino thoae findinga in gator detail and put thom into ppactivo with our pnt knowledge of the behaviour of

ZitO3. On loamLng to diotinguieb between a number of gion stimuli aniDolo tend to gene1ieo their ouboequant razyonooa to include oiDllcr cuao. Thus in the classic, ,; . oerinente of Pavlov (1927) dogs wero trained to salivate in response to a atiaulue that Preceded a reward of food. in other triala s, the dogs were poeonted with otinuli not followed by food. Pavlov found that not only did the pocitive otimulus elicit oa1ivation but other

eiDilar etinu1L did aleo. And the converse hold true for the negative btieulue and ones similar to it. After further trainings animals can be induced to mlan their genorslication and they say eventually bococo capable of

differentiating between air-Liar stimuli. An early example of such discrimination in quoted by Pavlov (op.cit.). Gubergrita b a student of bis D was able to train dogs to diotinguich between light grey and dark grey etimuli • Nora rocently s Skinner (1965) for

eeilo e has ehown that pigoono can learn to discriminate between of particular sL3os q whilot Sutherland (1969) baa &onetrated that various subjects (rats e goldfioh o octopuceo) can differentiate between cisilar obapoo or patterno. Different mathodo of training can ho uced in such 'diocrieivation..loozning' experiments

(fog' owmaW s aGo Switaloki et al. 196) 9 but Moste, at SOMO Stage involve the p000ntaticn of a ctiouluo associated ,with a negative reward. - 245 -

it is of obvious advantage for a pzdator to learn to distinguish between, say, a distasteful and palatable prey*' Hence inexperienced birds will take a wasp as readily as a fly, but soon learn to recognise the distinctive warning-colour pattern of the latter (1$2ostior 1935). On the other hand, predators should also benefit in being able to generalise on such noative stimuli. The evolution 'of Batesian mimicry is commonly regarded as arising through the agency of a combination of these two types of behaviour.

Various workers (e.g. M61mann 1934, Sexton 1959, Moz'x'eU and

Turner 1970, Duncan and Sheppard 1959) have investigated the degree to which a mimic must resemble a model before being mistaken for

It, But what implications have the above studies to the present work? They ahow that predators, including birds, are certainly capable of a high degree of specificity' in their response to conditioned stimuli. Yet it must be re-stressed that my experim3nts deal with types of equally-palatable prey • Those differ only with regard to stimulus U.-e. colour), not reward.

The apparently negative attributes of the unfamiliar typos can only be asociatod with their novelty, tie must now consider, in conjunction with the presônt rssulto, the evidence for discrimination learning and stimulus generalisation by predators searching for similar palatable prey. .One example illustrating the specificity of searching izeges for palatable prey has been mentioned in Chapter XX, Do Ruitor (1952) showed that caMd birds are capable of distinguishing between stick-catozi11are and the twigs of their host plant. The birds' were not totally adopt at recognising the caterpillars, however, for - 246 - soa of the twigs were apparently mistaken for prey and wore consequently pecked., t- hen other species of twig wore presented some of the birds pecked only those sticks that most resembled the larvae in size s shape and colour.

For the most relevant data. pertaining to discrimination on the basis of differences in colour alone we must again turn to the tt7ozc of crone (1967 9 1970)6 In his own words (page 43 9 1970) 9 the question he was asking was: "hew similar in terms of colour, form and atructure does a subsequent prey haw to be in order to elicit the response associated with the sample?" Here we are mainly concerned with the first type of visual cue. Croze first trained a pair of wild crows to search for mussel shells covering pieces of meat. Those prey were made by coating the shells with a layer of sand prior to being painted red. The experiment was carried out on a background of shingle.

After nearly two months 9 unrewarded test populations were occasionally presented. These consisted of 49 prey arranged in pairs.

A standard red mussel and a new shell comprised each pair and the two shells were separated by two metros • The now shells were distinguishable from the standard mussels by one of the three criteria mentioned above., namely: colour s, form or surface texture • All the prey-types were meant to be oszouflaged against the shingle. Colour variants were made from similarly shaped mussel shells as were used for the standard reds but were painted either a redder 9 yellower or bluer shade of rod e, or yellow or black • It is the results of tests

Involving these forne that are most portinent to the present discussion. The main findings were as follows. Though re&rad shells were - 247 -

turced as often as the standard redo the latter wore taken fitot more often from a pair, indicating that the birds were distinguishing between the two. Pad-yellows an the other hand wer* apparently indistinguishable from the standards • Blues were very different in appearance but also appeared more conspicuous; they were taken in smaller numbers • The yellow and blacks rezo totally overlooked because they were more camouflaged than blues and were equally different from reds • In other worth the crows' searching images were highly specific.

They tended to predate only those colours that were spectrally the most similar to the standard stimulus. Despite' the obvious differences in experimental design, my findings are coiuparablo with those of Cross. That is to say, the closer the resemblance of an unfamiliar prey to the familiar typo, the more likely is i to be taken. For axample o, the unfamiliar colours in the present training experiments were relatively less protected from predation than the unfamiliar colours of the grson'brown experiments described in Chapter V.

This will become obvious on comparing, for instance s the predation on the first day after each training session in the two groups of exporimsnts '(see Tables 12 IS and Tables 48a, b). Although the four khaki-training experiments should strictly not be directly compared, it may be relevant that the two most significant results were those for the first series (see x2 values, lowest row, Table t7) • t'5o 'may therefore infer that the birds were more liable to mistake 6's and a'° for 4 1s than they were 0's and 9"s. The results of experiment 4 shoved that the seven predatore, as a group, took relatively too many of the - 248 - familiar ''e from the populations presented at the start • In contrast to the ensuing decrease in the numbers of L&'s in the population, the proportions of Ei and 5 9s remained relatively static. The predators on masse pore apparently discriminating between k's and the two most similar colours. In fact when the data for individual birds were ewmimed we found distinct differences in behaviour, While all took an excess of k's, only one bird definitely tended to respond specifically to this colour. In most of the birds ensio brown preferences appeared to be interacting with the effects of training. Blackbird J, however, initially discriminated between 40 and 3 1 9, and 's and 5 1s. Hence the steep gradient, about a mean value of , of the curve shown in Figure l. This bird later tended to concentrate on searching for 's and in general he appeared to prefer the greener baits, Assuming that such a preference is representative of his normal behaviour we may conclude that during the start of the experiment it had been modified by the preceding period of training. In contrast to those for the other birds the graphs for J generally exhibit tall, sharp peaks around the modal frequency of Li's or

S's • In other words, his preferences, whatever their causes, wore very specific. In all the experiments the birds became loas specific in their choice as the experiments progressed. As they encountered new types of baits (by chance? of. Tinbergen 1960) they presumably learned that these too represented food.

Had the training periods in my oxperimnts been longer, the birds' tendencies to select the familiar colour may have been - 249 -

correspondingly more persistent, In Croze's (1967 1970)

experiments a 'new pair' Of crows arrived duzing the presentation

of the conditioned stimulus. Those birds had ostpez'ienos of the

standard reds for only about one woe!k lo whilst the 'older' pair fed on the sap1e for nearly two months. During the presentation of the toot populations 9 the 'old' pair responded more specifically

to the standard red mossol than did the 'now' pair.

(b) Disn&tive selection

Experiment 4 has again shown that a group of blackbirds and songthresheo can differ markedly in their individual responses. Mother result was that most of the birds, with one notable occopticn, preferred the browner baits once the effect of the conditioning to 4's had worn off. These two findings agree with the results from other eperimonts described in this thesis and have been discussed in Chapter VIII.

The effeot of the combined predation wan to produce a prey population with a greater variance and two nodes with the larger pask in the green half of the distribution. The selection therefore showed signs of being 'disruptive' 1 (2ather 1953) in the sense that the population had acquired two 'optimal through selection against inteznediete types • At least three factors contributed to give the selection its overall disruptive effect and those are given below.

1 Dobshansky (e.g. 1970) prefers the term 'diversifying'. - - 250 -

A. The prQ2ioU training on 49a resulted in the birds acquiring

33flCO fOK these balt5 e and Other morpho t32'3 1hO3fOnD at an advantage • Consequently solootion at the beginning of the epowLeut tioo 4kruptive, and epotatio, toovor each bird was bohoving in a GimLlar, mmmr to ito fellots. 13. the o2oots of the training diminished, moot of the birds tended to ooeiih for the brorner baits which moulted in the population boacning skewed ftuar& the grener end of the trum- The birds tozte now moro vsrible in behaviour. C. One bird (J) pxofoind L'° and Vo and was largely responsible for the obvious diffozouce botoen the final ralati'oly low ft,equencies of thoo oolorc3 In eoppariocm with the high propoticus of 7's and 'o. in vietl of the diveree behaviour of the individual birds it is vmllkely that the msuXto oben in FLgro 14 could be 4uilLcated in a zoat oxporLuent ezing Inctil birds. £cither (195) bee ehn that 61,sx-Vtiva 0010CUOM em e ndor

GaTtUa eandftiono t, give nine to ycoihint. This contention hoe boon oportod by mathematical GzaaminatLen of Go= apoeial oeeoc (see, for oeeplo; Levene 1953, tvino and WacArthur 196) 0, and by oerlaetal studies on Dogpbi in particular by Thodey end his colLeguso (e.g. Thoday end Bom 1)59, Giboon and Thoday 1962, 19(33) • snperiment 4 also seems to oppont iathor'o ergient. There are tto (often confused) mior aspects of dLoruptL3 selection. First, by discriminating against intodicto phonotye3, it say proite pslyohism and maintain the d.tstiuetu000 of the

mho. Sod, under certain cIrcLvmtan0W, it may be capable

Of tOLut!u1ng the palyCOl?phioo. - 25]. -

(i) Naintenango . of dietincneeo of

io know of one oituatio - in mitic co1oapo1y xpT?kiez -

tthor3 disruptive selection by pzcdito keeps the co'pho viea11y

distinct. This has boot been &monstratod by ctad.too an the tropical ewa11ctoil butoE'f1i0o t1io dardmus Brown and

P. ocan L. (C1a3 and Shopped 160 12 Sheppard 1958 9 1969 000 also Ford 1964 9br ziieti) • The fma1ee of thoec palatcW.c

species aro often po1yrphic and cadh morph mimics a paztict10

detasteful buttwf]y. Theo is thts a promium on the cloec wesomblance of the morphs to their appropriate modola 9 and inter-

mediate form tend to be eliminated by sit.-dopendont &toc. The results of eperimant 4 eueot that disruptive selection by pr3datore may also be czpable of prompting and maintainLnS the visual diatinctneso of non-mimetic morpho, bearing in mind that the combined effect of the three ecmpononta of the selection may have given a somewhat unrealistic result (sea abovo p. 250). The mast relevant paint is that once the birds had beaoo acaustozd to

searching fr 4 1o, they were inclined to continuo to do co when jresonted with nool1y-distributsd' populatiano. Varieties t7hich were beet similar to 4 1 s taza at the greatest advantage.

CormspendUSly g the variance of the population i cTeeeeed as the freqnay of 's diminished.

wCM t7oh 4am at a relative disadvontao boccueo it tiao effectively comon. It was not oliminotod free the population - at

least tiithin the period of study. One reason for this tOO that the effects of conditioning decreased with time as the birds diocoverod the other typos of bait. Elimination of morph 4 might hove acced

if0 for oamp1o 0 it had been izoro oonopicucua than the other

varieties. - 252 -

Cain eta. 1960 and Tumor 1961 hao in fact suggested that

disruptive selection by pradators could act to eliminate ccnapicuouo

phenotypes auth that the optima come to resethie particular o1eents of the bacgroid. Certainly a number of species do appear to exhibit such 'cryptic polytorphia& (Cain at al. op.cit.) iind

In many oases some rrpha are green and othexe, brot7n.

2ciny terrestrial onvironitsuta • for eernp10 grassland.: 9

woodInnds mixed hoaga 0 ate. 1, are essentially composed of a cotcplaxi medley of numarous shades of groan and brotrn • Those colours

In many polymorphic species inhabiting such habitats Cases in point ares the catoillars, pupae and i!iagoo of csrtain pidoptora for example the iras of caomatwn papiUcnaria L.

(Poulton 198k); various species of 1antidao (Edmunds 1972); the nymphs and isags of nereus 1crididaa (Byrne 1967 Owen 1966a,

personal observations); the earthwore A11o1obopora ch1oriti L.

(Satohall 1967); the East African molluscivorous land snail Edsxituling obos (Gibb •) (personal observations); nuoreus

tree.frgs, e.g. Hvla regii.la (Resnick and Janmson 1960 rious Hyp1ius species (Sck*itz 1967 9 1971); the chameleon Chamaleo

bltaentatus (Ogilvie and Owen 1962) • NCMy other examples could be mentioned.

Variation even in Copaaa is based on green versus bretn, for

the yellow morphe often appear green when the snail is in its shell.

Cain et al. (1960) were first to point out that the coz'phs of

C. nor&4, seem to resemble components of the background. Po1yozhic species inhabiting other typos of habitat also shoe such correlations between the colours of the morphs and the background. Safriel (1070) discovered that the morpha of the - 253 - intertidal gastropods !Le rita foraka4i, Rc1us and N.,poL Linn resemble thi colours of the pebbles in their environment.. A similar case has been made for the limpet Aceaea digitalis (Giosel

1970). The morphs of certain species therefore tend to resemble elements of the habitat and this might imply that they are kept visually distinct by disruptive predator selection. Glesel (1970) has obtained indirect evidence that such a situation exists in

Acmaea diitalis. It is doubtful tthether ouch of the selection in experiment 4 ias related to the diversity of the background because none of the nine varieties appeared to be the same colour as any particular component and they seemed equally conspicuous. However the experimental design could easily be modified to test the maintenance of the distinctness of 'cryptic morpho' • Such a line of enquiry tyould surely be profitable.

(ii) 4ajutenanco of veb1iti

IJather (1955) has emphasised that one of the conditions that must be fulfilled for disruptive selection to maintain a polymorphism is that each optimal phenotype should be an effective part of the environment of the others • Sex is an obvious example I male and female phenotypes must occur together in the same population.

In mimetic polymorphism the success of each morph depends on the frequencies of the others, for the commoner the mimic the more likely it will become associated with palatability. Likewise, in apostatic polymorphism s the selective value of each phenotype is assumed to be inversely related to its frequency. - 254 -

Polymorphism in a palatable species will be maintained by a

'disruptive' predator only if the selection is frequency-dependent.

Experiment 4 showed that the birds tended to take the common

(familiar) colour, thus affording protection to the rarer forms, and suggesting that the. selection had an apostatic component. However, frequency-independent brown preferences and the diversity of behaviour hindered the detection of apostatic selection during the experiment's later stages. It is clearly essential to carry out further experiments to test whether apostatic selection can indeed act on stictly pymorphic (as opposed to dimorphic) populations as is suggested by the work of Popham (1942, and discussed on pp. 42-44 of this thesis).

6. SUMMARY

To reiterate, the main findings were:

Searching images can be very specific. As in other experiments, blackbirds (with one exception)

and songthrushes preferred brown baits. The birds were very variable in behaviour. En masse, they acted in a manner that may be capable of

promoting polymorphism in a prey species. - 255 -

APPENDIX D TABLE 54

'Normal distribution' experiment ki data for predation over every three generations by blackbirds E, J, K, L, II, and two aongthrushes. -256 - Table

Numbers of baits. taken over every three (or two) generations by individual blackbirds (B, J, K, L, N) and the two songthrushes (Tts).

Generations 0 - 2

irc K .L 14 T's Total 0 o 0 0 0 0 2. 2 1 0 1 2 2 0 3 8 2 5 1 4 6 0 5 21. p33 14 3 9 6 2 8 42 10 23 6 10 1 11 61 o5 5 10 5 2 1 11 34 4 •o 1 3 1 8 17 7 0 0 0 1 0 0 1 8 0 0 0 1 0 0 1 Totals 38 38 27 31 5 48 197

Generations

Birds B J K L N T's 0 2 0 0. 1 0 0 3 1 2 1 1 3 0. 3 10 2 4 1 1 .4 0 1 11 3 7 6 9. 7 0 10 39 0 10 37 Fq 4 8 6. 6, .7 5 8 28 7. 7 1 3 54 '6 4 9 1 5 0 8 27 10 7 4 2 • 2 1 0 1 8 1 0 0 2 0 1 4 Totals 40 53 27 37 1 37 195

Generations 6 - S

Birds B J IC L N T's rilota 0 1 0 .2 1 1 1 6 1 3 0 0 3 2 12 20 .2 .12 28 2 6 2 •3 3 3 11 5 7 6 3 14 46 4 9 5 6 6 4 7 37 g 5 3 15 7 2 2 8 37 6 3 14 3 2 0 3 25 7 3 12 0 4 0 6 25 8 . 0 0 1 0 0 0 .1 Totals 1 39 53 29 27 14 63 1 225 Tab1e247cont , )

Gene rat br s_9-11

Birds B J K t M T's Totals 0 2 0 0 0 1 1 4 1 2 0 4 1 1 4 12 2 8 0 5 8 4 8 33 7 1 8 4 5 14 39 7 7 3 5 2 14 38 26 5 3 9 49 0 3 3 1 12 2 4 0 8 27 7 o 4 3 2 0 5 14 8 o 1 0 0 3 1 5 Totals 30 51 30 27 19 64 221

Generati OT is 12 - 14

Birds B J K L 14 T'S Totals 0 1 0 1 1 1 3 7 1 0 2 1. 2.3 3 11 2 3. 0 4 1 10 7 25 c 3 1 5 5 8 7 35 2 4. 3 8 7 6 30 6 10 60 0 5 25 5 9 6, 1 12 2 1 6 5 27 7 1 6 0 1 1 3 12 8 1 1 0 3 2 0 7 Totals 23 51 21 31 44 44 1 214

Generations 15 - 17

Birds B J K L N T's Totals 0 . 1 0 1 3 0 0 5 1 5 1 6 5 8 1 26 2 4 2 2 6 5 6 25 3 6 7 5 4 11 3 36 1 8 3 3 1 18 34 2 4 10 . 4 4 3 27 2 5 1 3 2 7 20 7 3 6 2 2 1 4 18 8 1 1 0 1 1 4 8 • Totals 27 40 24 31 32 45 199

Generations 18 - 19

Birds B • J K . L N T's Totals 0 1 0 3 2 1 2 9 1. 3 2 2 1 2 5 15 • 2 3 1 9 6 .3 5 27 3 6 2 10 3 3 3 27 0 3 0 1. 1 1 .6 2 11 3 2 2 4 24 6 0 2 2 2 0 4 9 7 3 4 5 2 0 0 14 8 2 3 0 1 0 0 6 • Totals 20 28 34 20 12 23 137 - 258 -

CHAPTER X GENERAL DISCUSSION AND CONCLUSION - 259 -

CHAPTER X GENERAL DISCUSSION AND CONCLUSION

The results have been discussed in full in each of the six proceding Chapters. This final section attempts to bring together the several lines of thought which the work has invoked. The findings will be used as a basis for discussing apostatic selection in natural populations and I shall stress likely areas of future research. Before we do this we should justify the experimental set-up as being realistic of a natural situation, first from the point of view of the predator and second from the point of view of the prey.

1. WILD PASSERINES AS PREDATORS OF POLYMORPHIC SPECIES

My experiments have shown that certain wild passerines, notably the blackbird Turdus merula, tend to behave in a manner that is capable of actively maintaining a colour polymorphism in

a prey species. It could be argued that the results are meaningless, simply because these birds do not normally eat polymorphic prey. However the evidence is to the contrary. The blackbird, for example, is very catholic in its choice

of diet (Witherby et al. 1938, Collinge 1941, Snow 1958). Its favourite items include earthworms and caterpillars (Snow 1958) and

both of these groups contain polymorphic species, some of which

are even dimorphic for green versus brown (see p. 252). In addition

the remains of snails have been found in the stomachs of

blackbirds (Collinge 1941) although they "lack the nervous

equipment for hammering [snails] against an anvil" (Morris 1954). - 260 -

It is probable that snails are taken when they are small and weak (Norris oi!Lc.tt.)'. This would include the young of many polymorphic species (Cej2aa&& AriantaD etc.). If blackbirds (end indeed other avian species) take appreciable nwnbore 2 of yog e.g. Cepaea, it is possible that apostatic selection acts early in the snail's lift-history and not later, as has, been generally supposed (Clarke 1962a,b). Despite the number of 7, papers twitten on paea, ecologically uo have but scratched at the surface • Other polymorphic species have been even less

Investigated. Even if the blackbird is not an important predator of polymorphic species it is closely related to one that is - the aonthruah. Both species are essentially ground-feeders and share a similar diet s with the obvious exception that the song*hrush alone has developed the ability to break large snails.

The behaviour of bait-eating blackbirds may have direct implications to the behaviour of the 'Cepsea-eating songthruah. The few songthrlushea encountered during the study certainly did not differ noticeably from blackbirds in their responses to the baits • Admittedly, 8ongthrush T (of the experiments in Chapters VII and VIII) had the strongest brown preference of all the individuals investigated, but this may have been a chance result (p. 158).

The other predetors in the experiments (see Table 1, p.52 ) also undoubtedly encounter polymorphic prey in 'nature' (e.g. for

Parue app, see the work of Den Boer (1971), described below, p.265 ).

1 1orrie (195) also points out that blackbirds steal large snails from thrushes, at the critical moment when 'hammering' has ended and the snail has been extracted. This method has little relevance in the present context. 2 !itness Gibb'e (1956) remarkable data for the quantities of Littorina sp. consumed by a single rock pipit 1nthu9 a pipoleaa. - 26]. -

Starlings and house sparrows featured prominently in some experiments • These birds are 'social feeders' and it is possible

that they are more efficient 'spostatic' predators than Turdus spp.,

because 'social facilitation' may ensure that nearly all the

individuals concentrate on the profitable,. common varieties (see pp. I8190).

2, BAITS AS NATURAL PREY

It could also be argued that because the prey were artificial

the birds responded to them in an artificial manner. However the

evidence, as outlined below, does not give Duch support to this

possibility. The main colours of the prey, green and brown, are

typical of many palatable animals s, including certain polymorphic

species.. It is possible that the birds were reacting abnormally

because the baits did not taste particularly 'animal-like' • The

baits however contained a high proportion of animal fat., In addition, J.N.H. Smith has informed me that blackbirds show no significant preference when given a choice between lard-and-flour

baits either containing or not containing a meat-extract ingredient. Although the baits were immobile, many polymorphic / species may also rely on immobility as a primary line of defence (see below) • In any case, the baits were not completely without

behaviour, because their spatial arrangements were altered

periodically.. The usual density of the baits, two per square metre, is certainly coepcwablo with the densities of natural polymorphic - 262 -

ymy g pulatiouo of certain species occur at oon higher doneitico (see bolo). (o) Casual obaervaticno suggested that the hunting patterns of blackbirds and thrushes searching for baits tore little different frc tshen searching for omoU insects or toms conclusion that is given support by the work of J.I2. Smith'

(personal coimunications)..

(f) lbst iortant e the birds wom feeding in their normal surcundings and ters not subjected to the potentially distorting rigours of captivity., -,

8. EVIDENCE FOR APOSTATIC SELECTION

Chapter II examined the then existing evidence for apoctatic selection and posed three basic questions concerning the evidenco for (a) conditioning to a specific prsy (b) specificity of conditioning, (c) opostatic selection. Some answers can now be provided in the light of the epaz'ir.ents.

Conditioning ton pocifio p

The keystone to the hypothesis of apostatic selection is the assumption that predators, under certain conditions • hunt iorpbe on the basis of specific searching Liegoa. Evidence has been presented (mainly in chapters v and IX) in support of thic contentions birds can be trained to search for familiar baits.

pecificity of conditioning The eporisento described in Chapter IX show that the searching isago can be very specific when birds are trained on one

1SDith's uork has no-s appeared w a D.Th1. Oftrd (1971) 9 but this has not boon seen by the author othtine of writing. - 263 - colour and then presented with a choice between the familiar and a similar unfamiliar colour.

(c) Aostatic. selection Proof of searching image behaviour, is not proof of apostatic selection.. If searching images are readily acquired for rare morphs selection will act to eliminate variability. Predators must preferentially search for common varieties.. The training experiments of Chapters V and IX showed that this occurs when birds are familiarised with one colour. All the birds acquired searching images specific to the familiar,, i .e.. 'common'. morph. But this is unlikely to happen in nature. Some individuals may build up searching images from chance encounter with rare varieties,, and this contention is supported by the results from the low-density

9:1 experiments (Chapters VI-Vu!). The responses of individual predators thus tend to be more diverse than might be expected and other factors in addition to the one mentioned may also contribute to the heterogeneity (Ch. VII!).

It is wrong to discuss apostatic selection on the assumption that individuals of a given predator species are homogeneous in their responses • This notion is implied in the writings of various workers (e.g. Clarke 1962a, Elton and Greenwood 1970, but not Cook

1965) • The important criterion is that the overall response of the predators should be apostatic, as seems to have been the case

In my 9*1 experiments. A further point needs to be stressed. In his original paper on the subject, Clarke 1962a, p.59) defines apostatic selection as selection by predators that confers an advantage to "phenotypes that stand out from the norm". He means selection that favours - 26& -

morphs which are numerically rare • This is an over-simplification s, for the situation may often be complicated by frequency-independent selection by predators • in my 9:1 experiments, the Tux'dus, specieS on average preferred browns, and Arnold (unpublished) found that his songthruahes also have frequency-independent preferences for certain morphs of Ceaea. Thus • for example, a morph that, is 'fairly common'. numerically may be 'very common' to a predator which for some reason has a strong preference for its colour-pattern. If apostatic selection were acting, this morph would be at a greater relative disadvantage than would be expected on the basis of its apparent frequency. Hence it is important that analysis for evidence of apostatio selection attempts to take account of frequency-independent preferences. Elton and Greenwood (1970) have provided two mathematical modei for estimating the apostatic component of selection in situations where data have been obtained from a number of populations containing different proportions of two morphs of a species A definition of apotatic selection should therefore stress that a Oraret morph is not necessarily numerically uncommon • For example: apostatic selection is selection that confers an advantage to phenotypes which to the predator appear to stand out from the effective norm,

The next logical step is to extend the research to observe how predators respond to natural polymorphic prey. In fact the work of Arniold (see p • 46 ) supports the argument that songth rushes prey apostatically on Cepaea. The birds were not observed - 265 - continuously but his results imply that they were using morph- specific searching images. Recently, Den Boer (1971) has obtained some promising results for captive tits Parus major and P. ater feeding on dimorphic (green versus yellow) caterpillars of the pine looper Bupalus piniarius (L.). He has shown that tits can be conditioned to green larvae such that in a choice situation this morph is preferred (compare with conditioning experiments, Chapter V). Since the green morph outnumbers the yellow by 1 to 500, Den Boer (opçit.) suggests that apostatic selection is maintaining the polymorphism, even though the yellow appear more conspicuous on pine needles, the normal habitat of the larvae. Den Boer's work is perhaps the first direct proof that predators hunt morphs by specific searching image. An axiomatic feature of a species showing apostatic polymorphism is that some or all of its major predators could hunt by searching image. At present, however, we do not know to ihat extent searching image behaviour occurs In the Animal Kingdom. My results and those of others (e.g. Tinbergen 1960; Croze 1970; Dawkins 1971a; Miller, in preparation) apply to birds, which are probably often important predators of polymorphic prey. Similar behaviour may occur in other vertebrates (reptiles and amphibia) possessing colour-vision, but these Classes have not yet attracted much attention. Colour-blind mammals may s9ect apostatically on the basis of tone, or even smell (see p. 38). There is some evidence (Landenberger 1968) that invertebrates (starfish) also tend to hunt for common prey and this would have important consequences if such selection were shown to act on polymorphic species. - 266 -

O'Donald (1968) has discovered that glow-worms are certainly capable of distinguishing between various morphs of Cpaca nemoralis. The possibility of apostatic selection by invertebrates

clearly warrants further investigation.

4 • CONDITIONS FOR APOSTATIC POLYMORPFISM

Accepting that the experimental design 'is-representative of

a natural situation, we can now use the results as a basis for

discussing the conditions favourable to polymorphism maintained

by apostatic selection. It should be immediately pointed out that proof of apostatic selection in a natural population does

not necessarily imply that the polymorphism is so maintained. The effects of frequency-dependent selection could be swamped by

selection that is independent of the frequenóy of the phenotypes. However, Cook (1965) has calculated that the fraction of individuals eaten by predators has only to be of the same order of magnitude

as a frequency-independent coefficient which is tending to eliminate the variability. It must also be realised that proof of apostatia selection does not rule out the possibility that other mechanisms are helping to maintain the polymorphism.

My experiments were not planned specifically with the aim of testing the conditions for apostatic selection. As we have seen, many of the properties of the predator-prey system were kept as constant as possible throughout the work • Nevertheless the research has yielded some promising clues that could be followed up in the future.

For our present purposes we shall assume that polymorphic species are liable to attack by predators that hunt by searching - 267 -

image. In a discussion of the conditions most likely to favour

apostatic selection we would like to know the relationship between

the apostatic disproportion' (Greenwood 1369) and the various properties of the prey.

The apostatic disproportion is a measure of the strength of

apoatatic selection. For a dimorphic species where morph A is

rarer than morph, and assuming that the morphs are equally

attractive (in terms of palatability, conspicuousness, etc..), the epostatic disproportion is the cross-product ratio:

density of A x number of B eaten 1 density.ofB x number ofA-eaten

This term will be greater than unity if selection is apostatic.

Its magnitude will partly depend on the relative frequencies of

the morphs; the rarer is morph A, the greater its relative

selective advantage and hence the greater the disproportion. It would be a straightforward matter to determine how the magnitude

of the apostatic disproportion is related to the relative frequencies of the morphs. The heterogeneity within my experiments forced ma to confine attention to the situation at. low densities when one

morph is nine times as common as the other. Future work should include the presentation of populations containing the two morphs in. various proportions.

The magnitude of the apoatatic proportion is likely to

depend not only on the low frequency of the raze morph but also on the properties of the species as a whole. Conditions for apostasis should be such that searching images are acquired for

common morphs and not for rex's morphs (Clarke 1962a). This means,

1cf, cross-product ratio used in 9:1 oxporirent, e.g* p* 104* - 268 -

in eMat s that the n1ov of i,idii ntoo6 per unit tiro by a given prodator ohould be neither too hiah nor tao Lcti. 12 the pr ozn3 ootoed too often, Searching LEQo am likely to be oaqaimd fbz ram vartotios mdj, In ddition

3eoion my be such ae to2cow a mixed diet * o4 ty thor000 tone to randozaeso (Tinboen 1960) • IL' the proy aro eono'cd too izo13r, ocarahlimS irgeo aro im.ikoly to be aaqjthed 2or any uh. The 'risk of oncountort w111 in pert dapend on the proportiap of the po&ito; for axamp1e Ito otato of wngoz, the a'idtoso of Lt hwting p3tten., its vioual acuity etc. These factors tsore not invotigated in the p000nt aperi!zentQ because the birds

O3 in the nOZe1 ourrovadingo nd it was not possible to observo vo'y ninr aspects of their b0havlour at the oao titre. Here tio ebaL. diecute the factors afgbetlngthe risk of onCotter iththj in torao of the proportles of the proi species as a thole i.e • the deity, cowplcuousneoa j, and palatability of they.

(3) fl3$!

Clate (39820) etgteeto that aoitatio PDXYMorphisa tlill be favovwad only wbon the prey iz3 9 fzArly uncorimnO ouch that rave norpha am oouito inf3qeIatly. At too high a denaity, ay rc?3u1t from oarchLn3 images bolnG acquired for ram t32'hO oiid at too 1OI a denoity MWMorghlbo may zalain ou1t becae searching icoe am not acquimd for oven Do=n vez'ioties. GreenuOed (1969) bee eivd at u!Ion by udng a Wi0ro Mthar-atical l, crent (Sao below). -269-

The investigation of the relationship between the apoatatic disproportion and density was outside the scope of my experiments. Apostatic selection was observed when the baits were at a density of 2 per square metre on grass and soil backgrounds. In the absence of other experiments we do not know whether this density is 'intermediate' or 1higb 1 i However, at some higher density,:seleotion must tend to randomness,, for at maximum density the rarer morphs were preferred (Chapter IV, see also Allen 1972).

It seems that at very high prey densities predators tend to behave in a manner that is capable of actively eliminating variability. It need hardly be pointed out that the relationships between apostatic, stabilizing selection and density could be easily studied with baits and wild birds.

There is some indirect evidence that the intensity of apoatatic selection in my 9:1 experiments would be exceeded if the work was repeated at densities higher than 2 baits per square metre. Owen (1965a, b) has collected samples of Liraicolaria martensiana from Uganda and has shown that polymorphism ( as measured by the number of morphs) is directly related to density.

The populations that were most polymorphic wore those containing over 100 snails per square metre. This implies, that apostatic selection is directly related to density, at least in the samples examined. It is possible that populations of very high densities would tend towards monomorphism (of. Clarke l962a) • Similar results have since been reported by Smith (1971) for the West

African bivalve Donau- , rugosus.

Greenwood (1969) has stressed the importance' of determining - 270 -

the relationship between the magnitude of the apostatic disproportion and the density of the prey. Owen (1965a 0 b) explains the Limicolaria results (see above) on the basis of the assumption that the relationship is variable such that the disproportion is highest at high densities. Greenwood (p.cit.) points out that the data can also be interpreted by postulating that the disproportion is for example independent of density. He refers to Holling (1965) who has produced a mathematical model, supported by experimental evidence 0 which predicts that the proportion of deaths due to a vertebrate predator will be highest at intermediate densities, and lowest at high and low densities. For simplicity the overall selection due to deaths from other causes may be assumed to be independent of density. It is also likely to be independent of frequency and will lead to decrease in variability-if acting alonei If the disproportion is independent of density, then apostatic selection will be relatively more important at intermediate densities than at

high or low densities. There is a corollary to the hypothesis that in palatable species selection by predators tends to promote pqlymorphism at intermediate densities and monomorphism at high and low

densities. The model equally applies to variation between two or more palatable monomorphic species. My brown and green baits for example could ho regarded as two separate species instead

of two morphs of the same species. At intermediate overall densities apostatIc selection may favour mutations that accentuate

the dissimilarities of colour-pattern so that the two species tend to diverge in appearance (Clarke 1962a 0 Holling 1965). From the training experiments in Chapter IX we can conclude that -271- these differences do not have to be very large, particularly if one species is relatively more common than the other. When both species are rare, searching images are unlikely to be acquired for either, but when the overall density of the two species is high, there will be "no necessary divergence and indeed there may be convergence" (Clarke 1962a, p. 62). In fact this idea had long been foreshadowed by Wallace (1889) who suggested that a scarce edible species can closely resemble a sympatric abundant edible species and thus gain some freedom from predation. Van Someron and Jackson (1959) have proposed a Similar explanation for parallelism of colour-pattern in African butterflies. The essence of such 'safety in numbers' (Van Soineron and Jackson, p.cit.) is that at high densities, predators quickly become satiated so that the proportion of prey destroyed decreases after a certain density has been reached (Molling 1965). The proportion eaten in any one species will therefore be less than in the absence of 'mimicry'. The results from the maximum density experiments (Chapter IV) suggest an additional mechanism for the evolution of such'mimicry'among' palatable species.

(b) Conspicuousness

The risk of encounter will partly depend on the 'effective density' of the prey species, that is the average distance from which a particular predator can detect one of the prey. The smaller the 'direct detection distance' (croze 1967, 1970) the lower the risk of detection. The properties of the prey that affect the direct detection distance can be discussed under the collective term 'conspicuousness' and will include such factors as coloration, - 272 -

size, behaviour and environment Many colour polymorphisms are 'cryptic' (Cain, King, and Sheppard 1960) in the sense that the colour-patterns of some or all of the morphs appear to match components of the background Concealing coloration will reduce the direct detection distance and so reduce the chances of a naive predator finding a given individual To be effective, concealing coloration must be associated with 'cryptic' behaviour (Cott 1940) to ensure that the prey remains motionless in situations where most likely to be attacked, only fleeing at the last possible moment9 if at all.

Many polymorphic species exhibit cryptic behaviour. Indeed the habit may be a prerequisite for apostatic selection. it is impossible to give a comprehensive list, but polymorphic snails (e.g. çp), treefrogs (e.g. yperolius), isopods (e.g. phaeroma), grasshoppers (e.g. Homorocoryphus) are all characterised by spending many of their diurnal hours motionless In situations where they are exposed to predators. Examples In birds also .1 It seem to fit the rule. Polymorphism is common, herons (Ardeidao) and it is well-known that these birds stand motionless for hours as

they wait for prey 1'0 The polymorphic bush-shrikes (Malaconotus app.) are notoriously skulking birds of the African bush (Mackworth-

Praed and Grant 1955 9 and personal observations).

Obviously if a prey species IS too concealed then the intensity of predation may not be sufficient to allow for àpo stat Ic polymorphisms. One would not expect to find such variation in organisms that spend all their time underground, for example. On the other hand

1Murton (1971c) has made a different Interpretation of polymorphism in herons but he does not satisfactorily explain how the variation is maintained. - 273 -

conspicuous palatable species which rely on mobility for their first line of defence are likely to be noticed, in the first instance, by their behaviour. Selection for apostasy is unlikely to occur under such conditions and this might explain why non-mimetic polymorphism is uncommon in such groups as butterflies and many birds. The conspicuousness of a prey species will also depend on the nature of the environment: the risk of encounter may be expected to be less in dense habitats than in open ones. In fact Cain and Currey (1968) have shown that thrush predation on a population of çpaea neinoralis is of greater intensity in tussocky grass than in neighbouring nettle patches. A dense habitat will lower the intensity of predation in a number of ways. First, it will act as a physical barrier to the predator, hindering movement. Second, it will render the prey difficult to see, because there are more distracting components in the background vegetation. Third, it will enable a greater area for dispersion of the population. If the prey are prone to climb (and Clarke, 1962a, has suggested that polymorphic snails may tend to climb more than monomorphic species) then predators will have to search in a vertical plane as well as In a horizontal plane. Taking the last factor Into consideration we may now realise that a density of 100 Limicolaria snails per square metre of thicket may in effect be a good deal Tlower The effect of coloration on the magnitude of apostatic selection could be investigated by repeating my low density g.:.i experiments with either more cryptic or more conspicuous morphs. Care would have to be taken to ensure that natural colour preferences are accounted for. An even more intriguing study would be to repeat the experiments with greens and browns but in more 'difficult' or more 'easy' environments - 274 -

The risk to a polymorphic species would be decreased if it

is 'slightly distasteful' (Clarke 1962a). As Clarke points out, Cepaea may be slightly distasteful, for thrushes eat the snails in large numbers only when alternative food is scarce (Sheppard 1951, Goodhart 1958, Wolda 1963, Davies and Snow 1965). The same remarks probably apply to other polymorphic snails and slugs. A species that is to distasteful may be expected to become monomorphic through the evolution Of an aposematic colour-pattern, but certain distasteful species are in fact polymorphic (e.g. the South American butterflies Heliconius melpoinene, H. erato and H. doris, Turner 1965, 1968; the African butterflies Danaus chrysious and Acraea encedon, Owen 1970). It is possible that even these so-called 'protected' species suffer enough deaths from predation to allow for the maintenance of variability by apostatic selection; although, as Turner (1968) points out with regard to Helloonius melpomene and

R. erato, other mechanisms may be responsible. In my experiments the palatability of the baits relative to alternative available natural food was not measured, but they were obviously very attractive because they were usually taken in

preference to other prey in the plots. Hence 'tasteful' prey were actually intrinsic to the achievement of apostatic selection. This does not detract from the possibility that apostasis might have been more pronounced had the baits been less palatable. Further experiments should be carried out with baits of varying degrees of

palatability. polymorphic There are in fáàt some,species which may be thought to be positively palatable to predators; obvious examples occur in the - 275 -

Acrididae, such as the African edible grasshopper, Homorocop nitidulus. But the important criterion is the palatability of the species relative to other prey items of the apostatic' predator. The species may be eaten but rarely if it is low on the predatoxs palatability spectrum.

5. TYPES OF APOSTATIC POLYMORPHISM

Non-mimetic colour polymorphism Shave long excited the attention of research workers. Yet most efforts have concentrated on the genetics and associated ecology of the organisms; the actual visual appearance of the variation has been neglected. There is much work to be done, for example, on the relative conspicuousness of the morphs and, of course, as to whether they can instill specific searching images in predators. There is also much to be gained from a study of colour-polymorphism on a comparative basis, but on a much wider scale than was attempted by Owen (1963b). I shall now endeavour to present a brief interpretation necessarily preliminary and speculative, of the kinds of apostatic polymorphisms that may exist in the Animal Kingdom, We would be on safest ground when considering a species dimorphic for green versus brown and known to be subject to attack from avian predators In fact, such prey animals are probably not uncommon and sonio examples have been given in Chapter IX (p. 252). TO illustrate the point, consider the green-brown dimorphism which occurs in the larvae of many Lepidoptera. These caterpillars usually live on trees and bushes and are - 276 doubtless eaten by various species of birds. Here, then,is a situation that is similar to the one in my 9:1 and training experiments. It is very tempting to suggest that apostatic selection maintains the dimorphisms of such larvae. That this may indeed be the case is supported by Den Boer's (1971) findings that birds can acquire searching images for the green and yellow morphs of the larvae of the pine looper (see above, p. 265).

On the basis of the results from Chapter VIII ('trimorphic' experiment, 2.3d) and Chapter IX ('normal distribution' experiments), we can extend the argument to cover green-brown polymorphism where more than two morphs are involved. This applies to many Orthoptera- for example, the grasshopper Homorocoryphus nitidulus has six

morphs, three based on modifications of a green body colour, and three based on brown (Owen 1965c, 1966a). Cepaea might also be regarded essentially as having green and brown morphs (p., 252). The, species mentioned above and on page 252 live in habitats

where green and brown predominate in the background and it is difficult to escape the conclusion that the morphs 'mimic' these components. This hypothesis, first mooted by Poulton (1890),

does not detract from the argument that apostatic selection maintains the polymorphism, but provides an additional method for maintaining

the discontinuity of the variation. This reasoning can be extended

to polymorphic species in other types of habitats where the morphs

also appear to mimic elements of the background (page 253). Clarke (1962a, 1969) and Owen (1963b, 1966a) have stated that

some morphs of certain polymorphic species (Cpaea and Limicolaria)

do not seem to resemble any colour pattern in the habitat. This belief may be partly due to a misunderstanding of the principles of - 277 -

• concealment 0 which have been so well detailed by Cott (1940). To the well-trained, scarcely naive human predator, a mid-banded a semi-streaked Limicolaria, or a yellow-striped S12haeroma

may indeed appear highly conspicuous on any background. But to, a lowly vertebrate predator this may not necessarily be the case.

We can do no better than to quote from Cott (p.cit., pp.4$-49)

• himself: ho refers to disruptive coloration:

"Even.,tbe simplest disruptive patterns tend to hinder • recognition and so make for concealment.. ,Fox example, the East African Rena adsperca a large frog whose general colour-scheme is a patch-work of subdued earthy- browns and olive-greens - wears on its back a conspicuous yellow stripe extending from the snout right along the middle of the back. • .So far from drawing attention to the animal, however, the effect of this stripe is quite the reverse • In the first place, the yellow line - which, of course, in itself bears no resemblance whatever to its wearer, but rather suggests to the casual glance a twig or blade of grass - stands out from the back and catches the observer's eye. Secondly, by providing a strong incident of colour, it serves to flatten by contrast the less-defined half-tones by which the real form of the frog is differentiated from its surroundings. In the third place, it serves to bisect the form of the frog, so that the eye of the enemy is actually presented with a configuration of two half-frogs - which look very different from one whole one - and as likely as not the brain behind the eye will fail to join them together in recognition'

In other words such a prey animal gains survival value by virtue of its conspicuous disruptive stripe (or Other disruptive

marking). The effect is enhanced when some of the component patterns harmoniso with the background while others strongly

contrast with it (Cott, pit.). Thus certain parts of the prey stand out whilst others merge with the background and

consequently the real appearance of the prey cannot be readily determined. I suggest that some of the so-called 'conspicuous morphs

of certain species are disruptively coloured. In fact disruptive - 278 -

coloration may occur frequently in polymorphisms. For instance, of the morphs of !phaeroma rugicauda (West 1964) I would class all but one (the uniformly coloured 'grey' morph) as disruptively patterned. Other examples abound in hemiptera, snails, grasshoppers and frogs. -

Disruptive coloration and apostatic selection may be compatible concepts. A contrasting disruptive pattern will render the prey inconspicuous to a naive predator. But if the predator accidentally detects the animal then it has only to latch on to the conspicuous component of the pattern in order to find more prey. The acquisition of the appropriate, searching image should then be rapid for, once recognised, the patterns tend to be very noticeable. Rarer phenotypes would then be favoured • Thus, through the agency of apostatic selection, a given environment may be able to support more than one type of disruptive pattern within a species. In addition some morphs (e.g. 'grey' of S,phaeroina rugicauda, certain varieties of Cp) may bear a 'general colour resemblance' (Cott 1940) to the background. FinaUy some morphs may be truly 'apostatic', . For even when disruptive coloration Is taken into account there are still some varieties which appear to be conspicuous in any background - for example, bright pink unbandeds of Cepaea (Clarke 1962a, 1969), certain morphs of Fhilaanus and Llmicolaria (Owen 1963b). Such forms may indeed be maintained within populations by virtue of their difference from the norm. In some species (e.g. Cepaea) only the minority of the morphs may be conspicuous; in other species (e.g. Limicolaria) the opposite may be the case. It will be realised that disruptive coloration effectively - 279 - increases the visible differences between the morphs. A predator hunting SRhaeroma rugicauda is likely to acquire a '!phaeroma -shap4' searching image only when it has become familiar with grays. At other times it is likely to be searching specifically for 'T-shaped black patches' (the pattern morph), 'narrow bright yellow stripes' (the yellow morph), 'wide brick-red stripes' (the red morph), etc.

The morphs therefore effectively differ in their shape, as well as colour. In some polymorphisms in non-mimetic species all the morphs are conspicuous The variation in many of these species has been described as 'massive' (Moment 1962). This occurs in animals such as the butterfly clam Donax variabilis, the brittlestar Qphiopholis aculeaton, the coolenterate Cerianthus americanus. In these species it is rare to find two individuals that look alike and Moment suggests that it is the variation per so which is adaptive because it makes it impossible for predators to form specific searching images.

Croze (1967,' 1970) presents experimental evidence in support of this view. However as mentioned previously such a situation could only have evolved through the action of apostatic selection: 'reflexive' polymorphism may be the final stage of apostatic polymorphism. It is possible that massively variable species tend to live in

relatively uniform habitats such as sandy beaches - this is certainly true of Donax. These environments bear few components

which can be mimicked. For a palatable species exposed to predators and unable to camouflage itself effectively, massive polymorphism

might be the only answer. Even from the above outline we can conclude that the types of ii

supposed apostatic polymorphism are many and varied. There are specie - 280 - which all the morphs are coloured for concealment, whether by general colour resemblance or by disruptive coloration. Other species possess morphs which seem to be genuinely conspicuous. In still other species all the morphs appear conspicuous and the diversity is often massive. These categories are by no means exclusive for there may be all types of polymorphism from 'cryptic' to conspicuous. The classification could be made more complicated by bringing in such factors as environment, degree of polymorphism, palatability, etc., but in the absence of facts there is little point in further speculation at present.. Colour polymorphism has yet to be studied on a comparative basis.

6. CONCLUDING .RIMARKS

In 1962 Clarke used the term 'apostatic polymorphism' for the first time. His choice of adjective may then have appeared somewhat presumptuous because the experimental evidence for apostatic selection was undeniably scanty. But the last ten years have seen the steady accumulation of data. The work in this thesis represents a concerted attempt to test apostatic selection. It has, I hope, provided strong support for the hypothesis. Other workers (e.g. Crozo 1970; Dawkins 1971a; Miller, in preparation; Shutt, unpublished) have obtained comparable results with birds and artificial prey. Nevertheless, as we have seen in the present Chapter, there is still much to be learned from birds and baits. The field studies of Arnold (unpublished) and Den Boar (1971) are promising pointers to future work and the time is ripe for more direct observations of predators feeding on natural polymorphic prey species. - 281 -

REFERENCES

ADAMKEWICZ, S.L. 1969. Polymorphism in the land isopod Arniadillium nasutum. Heredity 24 : 249-263.

ALLEN, J.A. 1967. Selection by wild birds when feeding on dimorphic artificial.prey. B.Sc. thesis, University of Edinburgh.

— ó 1972. Evidence for stabilizing and apostatic selection by wild blackbirds. Nature, 237 : 348-349, (SEE APPENDIX F).

• and CLARKE, B. 1968. Evidence for apostatic selection by wild passerines. Nature, 220 ; 501-502.(SEE APPENDIX E). ALLISON, A.C. 19564 The sickle-cell and haemoglobin C genes in some African populations. Ann. hum • Genet., 21 : 67-89.

ARNOLD ) R.W. 1971 ea nemoralis on the East Sussex downs and the nature of area effects. Heredity, 26 : 277-298. BEUKEMA, J.J. 1968. Predation by the threespLned stickleback (Gasterostzus aculeatus L.) : the influence of hunger and experience. Behaviour, 31 : 1-127.

BLEST, A.D. 1957. The function of eyespot pattern in Lepidoptera, Behaviour, U : 209-256. BROWER, J.VM Z. 1958. Experimental studieS of mimicry in some North American butterflies, Part 1. The Monarch, Danaus p1exipp, and Viceroy, Limenitis archippus arc4ppp Evolution, 12 32-47

o 1960. Experimental studies of mimicry. IV. The reactions of starlings to different proportions of models and mimics. Am. Nat., 94 : 2.71-282. BROWER, L.P., BROWER, J.VAN Z., and WESCOTT, P.W. 1960. The reactions of toads (Bufo terrestris) to bumblebees (Bombus amoricanorum) and their robberfly mimIcs (Mailophor'a bouioides). Am.Nat., 94 : 345-355. BYRNE, O.R. 1967. Polymorphism in the Australian Acrididae. 1. Inheritance of colour pattorns in the plague locust Chortoicetesterminifera. Heredity, 22 561-568. CAIN, A.J. 1953. Visual selection by tone of çpaea nemoralis L. J. Conch., 23 333-336.

, and CURREY, J.D. 1963. Area effects in Cepaea. Phil. Trans. R. Soc., B. 246 : 1-81. - 282 -

CAIN, A,J., and CURREY. J.D. 1968. Ecogenetics of a population of çpaea nemoralis (L) subject to strong area effects. Phil. Trans. R. Soc., B. 253 : .447-482.

KING, J.M.B., and SHEPPARD, P.M. 1960, New data on the genetics of polymorphism in the Snail Cepaea nemoralis L. Genetics, 135 : 393-411.

and SHEPPARD, P.M. 19513a, Natural selection in Genetics, 39 : 89-116.

and SHEPPARD, P.M. 1954b. The theory of adaptive polymorphism. Am. Nat., 88 : 321-326.

- , SHEPPARD, P.M., and KING, J.M.B.. 1968. Genetics of some morphs and varieties of Cpaea nemoralis. Phil. Trans. R. Soc., B, 253 : 383-396.

CAMIN, J.H., and EHRLICH, P.R. 1958. Natura11ect1on in water snakes (Natrix sipedon) on islands in Lake Eyrie. Evolution, 12 : 504-511.

CAPRETTA, P.J. 1969. Establishment of food preferences in Callus gallus chicks. Anim. Bohav., 17 : 229-231. CARTER, M.A. 1967. Selection in mixed colonies of Cepaea nemoralia and Cpaea hortensis. Heredity, 22 : 117-139.

• 1968. Area effects and visual selection in cpaea nemoralis and çpaea hortensis. Phil. Trans. R.Soc., B, 253 : 397-446.

CHANCE, M,R.A., and RUSSELL, W.MS. 1959. Protean displays: a form of allaosthetic behaviour. Proc.zool.Soc.Lond., 132,: 65-70:

CLARK, B. 1960. Divergent effects of natural selection on two closely-related polymorphic snails. . Heredity,l4 *323-44.

• 1961. Some factors affecting shell polymorphism in paeao D. Phil. thesis, Oxford University.

• 1962a.' Balanced polymorphism and the diversity of sympatric species. In Taxonomy and Geopy, ad. D. Nichols, Systematics Association, Oxford.

• 1962b. Natural selection in mixed populations of two polymorphic snails. Heredity, 17 : 319-345. - 283 -

CLARKE, B. 1968. Balanced polymorphism and regional differentiation in land snails. Ch. 13 in Evolution and Environment, ad. E.T. Drake, Yale University Prose, Now Haven and London.

• 1969. The evidence for apostatic selection. Heredity, 24 : 3147-352.

, and O'DONALD O P. 1964. Frequency-dopondent selection. Heredity, 19 : 201-206. CLARK, C.A., and SHEPPARD, P.M. 1960. The evolution of mimicry in the butterfly pi1io darlanus. Horedity, 14. : 163-173.

, and SHEPPARD, P.M. 1962. Disruptive selection and its effects on a metric character in the butterfly P2pilio dardanus. Evolution, 16 : 214-226, COLLINGE, W.E. 1941, The food of the blackbird (Turduo marula Ii.) in successive years. Ibis, 5 : 610-613. COOK, L.M. 1965. A note on apostasy. Heredi!, 20 : 631-636.

, 1967. The genetics of çpaea nemoralis. Heredit y , 22 :"397-410.

1971. Coefficients of NCtural Selection. Hutchinson, London.

COTT, IhB. 1940.Adaptive Coloration in Animals. Methuen, London, CROWLEY, T.E., and PAIN, T..1970. A monographic revision of the African land snails of the genus Limicolaria Schumacher (Moliusca - Aohatinidao). Mu .R .Afriquo Centr. . Zool., 177 : 1-61. CROZE, H.T. 1967 Dispersion in camouflaged animals and searching image in predators. D.Phil. thesis, Oxford. 1970. Searching image in carrion crowa. Z. Tierpoho3..., 5:1-85. DAVIES, P.W., and SN0tS, D.W. 1965. Territory and food of the song thrush. Br.Biz'ds, 58 :, 161-175. DAWNINS, C.R. 1966. Selective feeding in the domestic chick. D. Phil. thesis, Oxford UnivOraity. DAWKINS, U. 1971a. Perceptual changes in chicks: another look at. thQ 'search image' concept. Anim.Behav,, 19 : 566-57,46 1971b. Shifts of 'attention' in chicks during feeding. Anim..Bohav., 19 : 575-582. - 2814 -

DEMPSTER, E.R. 1955. Maintenance of genetic heterogeneity. Cold Spring Harb. Syn. qnt. Biol., 20 : 25-32.

DEN BOER, N.H. 1971. A colour-polymorphism in caterpillars of pa1us piniarius (L.) (Lepidoptera: Geometridae) Neth. J. Zool., 21 : 61-116.

DIVER, C. 1929. Fossil records of mendelian mutants. Nature, 1214 : 183.

DOBZHANSKY, Th. 1951. Genetics and the Origin of Species, 3rd ed. Columbia University Press, New York.

• 1970. Genetics of the Evolutionary Process. Columbia University Press, New York and London. DUNCAN, C.J., and SHEPPARD, P.M. 1964. Senbry discrimination and its role in the evolution of Batesian mimicry, Behaviour, 214 : 269-282. ECK, P.J. VAN. 1939. Farbensehen und Zapferfunkt ion bel der Singdrossel Turdus e. ericitorum. Arch. N4erl. Zool., 3 : 1450-1499. EDMUNDS, N. 1972. Defensive behaviour in Ghanaian praying mantids. Zool. J. Linn. Soc., 51 : 1-32. ELTON, R.A., and GREENWOOD, J .J .D. 1970. Exploring apostatic selection. Heredity, 25 : 629-633. ESKER, G.C. 1968. Colour markings in Phacops and Greenops from the Devonian of New York. Palaeontology, 11 : 1498-500.

EWERS, W.H. 1966. Polymorphism in fossil Velacumantus australis (Gastropoda: Pomatidldae). Mist. J. Zool., 114 : 867-875. FISHER, R.A. 1930. The GeneticalTheory of Natural Selection. Clarendon Press, Oxford.

and YATES, F. 1963. Statistical Tables for Biological, Agricultural and Medical Research, sixth edition. Oliver and Boyd, London.

FOERSTE, A.F. 1930. The colour patterns of fossil cephalopods and brachiopods, with notes on gasteropods and pelecypods. Contr. Mus. Palaeont. Univ. Mich., 3 : 109-150. FOIñ'AINE, A.R. 1962. The colours of Qphiocemina nigra (Abilgaard), 1. Colour variations and its relation to distribution. J. mar. Biol. Ass. U.K., 42 : 1-8. FORD, E.B. 1940. Polymorphism and taxonomy. In The New Systematics, ed. J. Huxley, Clarendon Press, Oxford. - 285 -

FORD, E.B. 1953. The genetics of polymorphism in the Lepidoptera. Adv. Génet., 5 43-87. .1964. Ecological enetics. Methuen, London.

1965. Genetic o1phiam, Faber and Faber, London. FOX, D,L., and PANTIN, CF.A. 1941. The colours of the plumose anemone Metridium senile. 'Phli.Trans. R. Soc., B, - .20 ; 415-450. FØYN, B., and GJØEN, 1. 1954. Studies on the serpulid Pomatoceros tqieter L. II. The colour pattern of the branchial àx'own and its inheritance. tt. Mag. Zool., 2 85-90.

GIBE, J.A. 1956. Food, feeding habits and territory of the Rock Pipit Anthupino1etta Ibis, 98 506-530. • .1962. L. Tinbergen's hypothesis on the role of specific search images. Ibis, 104 : 106-111. GIBSON, J.B., and THODAY, J.M. 1962. Effects of disruptive selection. VI. A second chromosome polymorphism. Heredity, 17 1-26. , and THODAY, J.M. 1963. Effects of disruptive selection. VIII Imposed quasi-random mating. Heredity, 18 : 513-524. GIESEL, J.T. 1970. On the maintenance of a shell pattern and behaviour polymorphism in Acmaea digitalis, a limpet Evolution, 24 : 98-119. GILLESPIE, T.H. 1932. A book of King Peguns. Jenkins, London. GOIN, C.J. 1950. Colour pattern inheritance in some frogs of the genus E1euthrodacyB. Bull. Chiçgo Acad. Soi1, 1 : 1-15. GOODHART, C.B.1958, Thrush predation on the snail Cepaea hortensis, J. Aniin. Ecol,, 27 : 47-57. GORDON, H., and GORDON, M. 1957. Maintenance of polymorphism by potentially Injurious genes in eight natural populations of the platyfish, Xiphophorus maculatus. J, Genet., 55 1-44. GOSS-CUSTARD, J.D. 1970. The responses of redshank Tring totanus to spatial variations in the density of their prey. J.Anim.Ecol., 39 : 91-115. GREENWOOD, J.J.D. 1969. Apostatic selection and population density. Heredity, 24 157-161. - 286 -

HALDANE, J.B.S. 1955. The biochemistry of heterosis and the stabilization of polymorphism. Proc.R.Soc. ,B, 144 : 217-220. - , and JAYAKAR, S.D. 1963. Polymorphism due to selection depending on the composition of a population. J.Genet., 58 : 318-323. HALKKA, 0., and LALLUKKA, R. 1969 - origin of balanced polymorphism in the spittlebugs (Philaenus, Homoptera). Ann. Zoo].. Fenn., 6 : 431-434. HALL, B,P., MOREAU, R.E. and GALBRAITH, I.C.J. '1966. Polymorphism and parallelism in the African bush shrikes of the genus Malaconotus (including Chiorophoneus). Ibis, 108 161-182. HAYES, J.T. 1962. Studies of the searching image in the eastern black-capped chickadee. B.Sc. thesis, Axnhart College. HEPPNER, F. 1965. Sensory mechanisms and environmental clues used by the American robin in locating earthworms. Condor, 67 27-255. HOLLING, C.S. 1965. The functional response of predators to prey density, and its role in mimicry and population regulation. Mom. Ent • Soc. Canad., 45 : 1-60. HUNPHRIES, D.A., and DRIVER, PJ. 1967. Erratic display as a device against predators. Science, 156 1767-1768.

• - ,'and DRIVER, P.M. 1970. Protean defence by prey , animals. Oecologia, 5 : 285-302.

JOHN, ER.,• ' CHESLER, P., BARTLETT, F., VICTOR, I. 1968. Observation learning in cats. Science 9 159 : 1489-1491. KATZ, D. 1937. Animals and Men. Studies inomparative ,Sychology. Lóngmans, Green aud Co., New York. KEAR, J. 1962. Food selection in finches with 'specific reference to interspecific differncea • Proc. zool. Soc. Lond, 138 ; 163-205. .1964. Colour preferences in young Anatidae. Ibis, 106 '361-369. KETTLEWELL, H.B.D. 1955. Selection experiments on industrial melanismin the Lepidoptera. Heredj, ! : 323-342. 1958. A survey of the.frequencies of BistOñ betularia (L.) (Lep.) and its melanic forms in Great Britain. Heredity, 12 : 51-72. - 287 -

KLOPFER, P.H. 1961. Observational learning in birds. Behaviour, 17 : 71-79.

KONTS, N. 1923. tJntersuchungen Uber die rkenntr, des Schipansen. Zoopsychologisc des Museum Darwinianum,. Moscc,w,

LACK, D. 1954. The Natural Regulation of Animal mumbers. Oxford University Press. LANDENBERGER, D.E. 1968. Studies on selective feeding in the Pacific starfish Pisaster in southern California. Ecology, 49 : 1062-1075. LEVENE, H. 1953. Genetic equilibrium when more than one ecological niche is available. Am.Nat., 87 : 331-333. LEVINS, R., and MACARTHUR, R. 1966. The maintenance of genetic polymorphism in a spatially heterogeneous environment: variations on a theme by Howard Levene. Am. Nat., 100 : 585-589. LI, C.C. 1962. On 'reflexive selection'. Science, 136 : 1055.

LORENZ, K. 1962. On the function of colour in coral reef fishes. Proc. R. Xnstn. Gt. Br., 39 : 282-296, MACKWORT!i-PRAED, C.W., and GRANT, C.H.B. 1955. Birds of Eastern and North Eastern Africa. Volume 2, Longmans, London. MARPLES, G. 1933. Experiments on colour-sense in birds. Br.Birds., 26 : 238-242.

MATHER, K. 1953. The genetical structure of populations. Soc. exp. Biol,, 7 : 66-95. .1955. Polymorphism as an outcome of disruptive selectbn. Evolution, ! : 52-61. MOMENT, G.B. 1962. Reflexive selection: a possible answer to an old puzzle. Science, 136 : 262-263. MOK, J.H., MOOK, L.J., and HEIINS, H.S. 1960. Further evidence for the role of 'searching images' in the hunting behaviour of titmice, Arch. Néerl. Zool., 13 448-465.

MORRELL, G.M. and TURNER, J.R.G. 1970. Experiments on mimicry: the response of wild birds to artificial prey. Behaviour, 36 : 116-130. MORRIS, D. 1954. The snail-eating behaviour of thrushes and blackbirds. Br.Birds, 47 : 33-49. 288 -

HOSTLER, G. 1935. Beobohtting zur 1,vage der Wespenmimikry. Z Morph. Okel. Tiere, 29 : 381-454. r4IJHLMANN; H. 1934. 1w Modeliversuch Kunstlich erseute Jiimikry

und ihre Bedeutung fUr den 9Nachahmer'. Z. Morph. Okel. Tiere, 28 ; 259-296. MULLER, F. 1878. (Notes on Brazilian entomology). Trans.ent. Soc. Load., 3 211-223. The Munsefl. Book of Color. (Munsell Color Co.Inc., Baltimore, 1966). MURTON, R.K. 1971a. Why do some bird species feed in flocks? Ibis, 113 : 5313-536. .1971b. The significance of specific search image in the feeding behaviour of the wood-pigeon. Behaviour, 40 : 10-132. .1971g. Polymorphism in Ardeidas. Ibis, 113 : 97-99. O'DONALD, P.M. 1968, Natural selection by glow-worms in a population of Cepea nemoralis. Nature, 217 : 1913. OGILVIE, P.W., and OWEN, D.F. 1964. Colour change and polymorphism in Chameleo bitaeniatus. Nature, 202 : 209-210. OWEN, D.F. 1963a. Polymorphism and population density in the African land snail Limicolaria martensiana. Science, 140 : 666-667. .1963b. Similar polymorphisms in an insect and a land snail. Nature, 198 : 201-203. -1965a. A population study of an equatorial land snail Limicolaria inartensiana. Proc • zool, Soc. Lond., 1413 361-382. .1965b. Density effects in polymorphic land snails. Heredity, 20 : 312-315.

.lgSSc. Swarming and polymorphism in the African edible grasshopper Homorocoryphus nitidulus (Tettigonioidea, Conocophalidae). Acta Trop., 22: 55-61.

.1966a. Animal ology in Tropical Africa. Oliver and Boy4, Edinburgh and London.

•1966b. Polymorphism in Pleistocene land snails. Science, 152 : 71-72. - 289

OWEN, D.F. 1967. The interpretation of a polymorphism in the African bush shrikes. This, 109 : 278-279.

41970. Mimetic polymorphism andthe palatability spectx'um. Olkos, 21. : 333-336.

, and WEIGERT, R.G. 1962. Balanced polymorphism in the meadow spittlebug, Philaenus spumarius. Am. Nat., 96 : 353-359.

PAVLOV, I.P. 1927. Conditioned Refle xes (translated by G.V. Anrap). Oxford University Press, London.

PAYNE, R.B. 1967. Interspecific communication signals in parasitic birds. Am. Nat., 101 : 363-376. PEARL, R. 1911. Data on the relative conspicuousness of barred and self-colored fowls. Am. Nat., 45 : 107117

PIELOWSKI, Z. 1959. Studies on the relationship: predator (goshawk) - prey (pigeon). Bull.Acad. po1. Sd. Cl. II Ser, Sci. biol., 7 : 401-403.

• 1961. Uber den Unifikationseinfluss der selektiven Nahrungswahl des Habichts (Accipiter gentilis) aus Haustaben. 9 : 183-194. POPHAM, E.J. 1941. The variation of the colour of certain species of Arctocorisa (Hemiptera Corixidae) and its significance. Proc. zool. Soc. Lond., 111 135-172.

• 1942. Further experimental studies on the selective action of predators. Proc. zool. Soc. Lond., 112 105-117.

POUGH, F.H. 1964. Apostatic selection and the searching image. B.Sc. thesis, Amherst College,

POULTON, E.B. 1884. Notes upon, or suggested by, the colours, markings, and protective attitudes of certain lepidopterous larvae and pupae, and of phytophagous hymenopterous larvae. Trans. ant. Soc.Lond., (1884) 15-56.

• 1890. The Colours of Animals, Kegan, Trench, TrUbner and Co., London.

PROP, N. 1960. Protection against birds and parasites. Arch. Nev1. Zool., 13 : 380-445. - 290 -

PUMPHREY, R.J. 1948. The sense organs of birds. Ibis,, 90 171-190.

.1961. Sensory organs: vision.. In Biology, and comparative physiology of birds ., vol. II, ed. A.J.Marshall, 55-68, Academic. Press, New York.

RABINOWITCH, V.E. 1961. A study of the effect of early e,perience on stereotyped food habits in the domestic chicken. M.S. thesis, University of Wisconsin.

• 1968. The role of 'experience in the development of food preferences in gull chicks. Anim. Behav., 16 : 425-428. REIGHARD, J. 1908. Warning coloration in coral-reef fishes. Publs.Carnegie Instu, 103 : 257-325. RESNICK, LA., and JAMESON, D.L. 1963. Colour polymorphism in Pacific tree frogs. Science, 142 : 1081-1082.

RUITER, L. DE. 1952. Some experiments on the camouflage of stick caterpillars. Behaviour, 4 : 222-232. SAFRIEL, U. 1969. Ecological segregation, polymorphism and natural selection in two intertidal gastropods of the genus Nerita at Elat (fled sea, Israel). Israel J. Zool., 18 : 205-231., SALT, G.W. 1967. Predation in an experimental protozoan • population (Woodruffia-Paramecium). Ecol . Monogr., 37-: 113-144. ' SATC}LL, J.E. 1967. Colour dimorphism in Al1o1opa chloritiøa Say. (I1umbricidae). J. Anim. Ecol., 36 623-630.

SCHIØTZ, A. 1967. The treefrogs (Rhacophoridae) of West Africa. po1ia zool. Hue. haun., 25 : 1-346,

• 1971. The supergenus yperoiius viridiflavus (Anura). Videnek. Meddr dank naturh. Foren., 134 : 21-76.

SCHMALHAUSEN, I.I 1949. Factors of E.volution: the Theory of Stabilizing Selection. Blakiston, Philadelphia.

SELOUS, F.C. 1908. African Nature Motes and Reminiscences. Macmillan and CO., London.

SEXTON, O.J. 1959. Experimental studies of artificial Batesian mimics. Behaviour, 15 : 244-252. - 291 -

SHEPPARD, P.M. 1951. Fluctuations in the selective value of certain phenotypes in the polymorphic land snail Re2aea nemoralis (L.). Heredity, 5 :125-134.

• 1958. Natural Selection and Fiieredity. Hutchinson and Co., London.

• - . 1969. Evolutionary genetics of animal populations: the study of natural populations. Proc. XII Int. Conf. Genet., 3 :261-279. SIMPSON, C.C. 1953, The Major F'eatures of Eolut1on. Columbia University Press, New York. SKINNER, B.F. 1965. Stimulus generalisation in an operant: an historical note. In Stimulus Qreneralisa ton, ed. D. Mostofaky, 193-209, Stanford University Press, Stanford. SMITH, D.A.S. 1971. Polymorphism and population density in Donax rugsus (Lemellibranchiata: Donacidae), 3. Zool., Lond.•, 164 : 428-1+42, SMITH, J. MAYNARD. 1962, DIsruptive selection, pLymorphism and sympatric speciation. Nature, 195 : 60-62.

SMITH, J.N.M. 1961. Experiments on the food 'selection of the minnow (Phoxinus phoxinus L.). B.Sc. thesis, University of Edinburgh.

1971. The searching behaviour and prey recognition of certain vertebrate predators. D.Phil.Thesis, University of Oxford. SNOW, D.W. 1958. A Sttudy of B;lackbirds. Allen and Unwin, London. SOANE, I.D. 1970, Mouse predation and its possible effects on polymorphic prey. B.Sc. thesis, University of Edinburgh. SOMERON, V.G.L. VAN, and JACKSON, T.H.E. 1959. 'Some comments on protective resemblance amongst African Lepidoptera (Rhapalocera). J. Lepid. Soc., 13 : 121-150. SUMNER, F.B. 1934. Does 'protective coloration' protect? - Results of some experiments with fishes and birds. Proc. natn. Acad. Sd. U.S.A., 20 559-564. - 292 -

SUTHERLAND 9 N.S. 1969. Outlines of a theory of visual pattern recognition in animals and man • In Animal iscrimnation Learning, ed. R.M. Gilbert and N. S. Sutherland 385-411 Academic Press, London and New York.

SWITALSKI, R.W. 0 LYONS, V., and THOMAS, D.R. 1966. Effects of interdimensional training on Stimulus generalisation. J. op. Pscho1,, j : 661-666. THODAY, J .M. and BOAM, T.B. 1959. Effects of disruptive selection. II. Polymorphism and divergence without isolation. Heredity, 13 : 205-218. THORPE, N.H. 1956. Learning and Instinct jnnimals. Methuen, London. TINBERGEN, L. 19146. De sperwer als roofvijand van songvogels. Ardea, 34 : 1-213.

o 1960. The natural control of insects in pine woods, I. Factors influencing the intensity of predation by songbirds. rch.Ner1. Zool., 13 : 265-343. TINBERGEN, N., IMPEKOVEN, M., and FRANCK, D • 1967 • An experiment on spacing-out as a defence against predation. Behaviour, 28 ; 307-321. TOLMAN, C.W. 1968. The varieties of social stimulation in the feeding behaviour of domestic chicks. Behaviour, 30 : 275-286. TURNER, E.R.A. 1961. Survival values of different methods of camouflage as shown in a model population. Proc. zool, Soc. Lond4, 136 : 273-283.

-_. 1965, Social feeding in birds. Behaviour, 214 1-143, TURNER, J.R.G. 1965. Evolution of complex polymorphism and mimicry in distasteful South American butterflies. Int. Co. Entom., 12 : 267.

1968. Some new Heliconius pupae: their taxonomic and evolutionary significance in relation to mimicry (Lepidoptera, Nymphalidae). J. ZoO1.,4., .155 311-325. UEXKULL, J, VON. 1934. Steifgeduroh die mwelten von Tieren und Menschen. Springer, Berlin.

VERNON 9 M.D. 1957 • A Further study of visual ecepti. Cambridge University Press, New York. - 293 -

WADDINGTON, C.H. 1957. The Strategy of the Genes. Allen and Unwin, London. WALLACE, A.R. 1889. Darwinism: an Exposition of the Theory of Natural Selection, with Some of its Applications. Macmillan and Co., London.

WALLS, G.L. 1942. The Vertebrate Eye and its Adaptive Radiation. Cranbrook Press, Bloomfield Hills, Michigan.

WARD, P. 1965. Feeding ecology of the black-faced dioch Quelea Quelea in Nigeria. Ibis, 107 : 173-214.

WEIDMANN, U. 1965. Colour preference and pecking response in young moorhens and coots. Ibis, 107 : 108-111.

WEST, D.A. 1964. Polymorphism in Sphaeroma_gicauda. Evolution, 18 671-684. WILLARD, D.E. 1966. The thermoecology of CnemidophorustigVj. Ph.D. thesis, University of California, Davis.

WILLIAMSON, M.H. 1958. Selection, controlling factors and polymorphism. Mi. Nat., 92 : 329-335. WITHERBY, H.?., JOURDAIN, F,C.R., TICEHURST, N.F. ,. TUCKER, B.W. 1938. The Handbook of British Birds. H.F. and G.W.Witherby, London. WOLDA, H. 1963. Natural populations of the polymorphic landsnail qpaea nemoralis (L.). Arch. Ner1. Zool., 15 381-471. WRIGHT, S. 1948, On the role of directed and random changes in gene frequency in the genetics of populations. Evolution, 2 : 279-294. WYNNE-EDWARDS, V.C. 1962. Animal Dispersion in Relation to Social Behaviour. Oliver and Boyd, Edinburgh.

YOUNG, P.T. 1948. Studies of food preference,. appetite and dietary habit. VIII. Food-seeking drives, palatability and the law of effect. J • comp. physiol. Psychol., 41 : 269-300. ZAHAVI, A. 1971. The function of pre-roost gatherings and communal roosts. Ibis, 113 : 106-109. - 294 -

APPENDIX E 4 PUBLISHED PAPERS

ALLEN, J.A. and CLARKE, B. 1968. 'Evidence for apostatic selection by wild passerines'. Reprinted from Nature, 220: 501-502.

AND

ALLEN, J.A. 1972. 'Evidence for stabilizing and apostatic selection by wild blackbirds'. Reprinted with corrections of publisher's errors, from Nature, 237: 348-349. - 295 -

(Reprinted from Nature, Vol. 220, No. 5166, pp. 501-502, November 2, 1968)

Evidence for Apostatic Selection by Wild Passerines IT has been suggested that predators concentrate on common varieties of prey, and tend to overlook rarer forms even if they are obvious" 2. Within a prey species this apostatic selection could act to maintain a balanced

polymorphism2 ' 3 . The selection would then be dependent on frequency in the sense that the selective value of each

phenotype would vary inversely with its frequency 4 . 5 . Direct evidence of such selection has been obtained from experimental studies of predation by fish 2, and there is

indirect evidence of similar sehiction by birds 1,2 ' 6 . Ground- feeding passerines are important predators of many polymorphic insects' and molluscs, and we report here some preliminary experiments designed to investigate their behaviour when presented with artificial dimorphic prey'. We used wild birds in their normal surroundings. The most frequent species were blackbirds (Turdus merula L.), house sparrows (Pa88er domesticu.s (L.)), dunnocks (Prun- ella modula.ris (L.)), starlings (Sturnu8 vulgari8 L.) and robins (Erithacu8 rubecula Hart.). The prey were cylindrical baits (07 cm long x 07 cm in diameter), made of flour and lard in the proportion 5 2 by weight. They were coloured green or brown by the addition of edible dye (11 cm3 dye to 1,000 g of mixture). In other respects, including hue and brightness 8, the baits were apparently identical. We find no evidence of differences in taste. The first series of experiments involved populations of baits in which one colour was nine times commoner than the other. In August 1966 a square grid of 100 one- metre squares was marked out on the lawn of the Depart- ment of Zoology at EdinburgI. Within it, 180 green and twenty brown baits were distributed at random, two baits in each metre square. To the human eye the browns were slightly more conspicuous. The spatial distribution of baits was changed each day. During the experiment the grid was under continuous observation, and a record was kept of the numbers of baits taken by the birds at each visit. Frequent replacement of baits ensured that the 9 : 1 ratio was kept constant. After seven days of predation, the experiment was repeated with a population of 20 greens and 180 browns. The daily totals of baits eaten are shown in Table 1. Below the grand totals are given the expected numbers based on the assumption that there is no selection. The blackbirds and starlings took more browns than expected in both experiments. Dunnocks and house sparrows, on the other hand, took more greens in the first experiment and more browns in the second. It is clear from these results that the birds exercised visual selection, - 296

Table 1. DAILY TOTALS OF BAITS TAKEN: GRASS ]BACKGROUND Experiment 1 180 greens : 20 browns Hours of Observed predation - Date observation Blackbirds Starlings Dun- H. Sparrows (BST) (8+) (2) nooks (2) (many) - U B B GB U B - 30.7.66 9.03-12.00 2 0 1.8.66 10.30-18.15 61 9 0 1 1 0 - - 2.8.86 8.30-16.30 71 22 - 4 0 - 3.8.66. 9.20-15.15 71 17 -. - 5 0 10 1 5.8.66 10.30-17.05 45 11 - - 14 0 86 2 6.8.66 5.09- 7.51 69 15 - - - - 14 0 7.8.66 16.06-18.10 35 12 20 2 - - 57 0 Grand totals 354 86 20 3 24 0 167 3 Expected 9 : 1 3960 440 20.7 2-3 21-6 2-4 1530 17-0 Expected, constant selec- 349-0 910 15-2 7-8 23-4 0-6 154-4 15-6 tion- The approximate numbers of birds involved in each experiment are given at the heads of the columns. Table 1. (Cont.) Experiment 2 20 greens 180 browns Hours of Observed predation observation Blackbirds Starlings Dun- H. sparrows (BST) (2+) (many) flocks (2) (many) Date U B U B GE U B 9.8.66 9.30-15.15 - - 0 7 0 1 15 42 10.8.66 10.30-17.30 0 5 - - - 5 155 11.8.66 10.50-16.10 0 14 1 37 - - 14 183 14.8.66 15.05-18.10 0 13 0 45 0 2 4 136 15.8.66 11.40-14.10 0 13 0 44 - - 2 60 17.8.66 14.30-18.35 1 11 1 . 19 - - 9 74 18.8.66 10.25-17.40 0 24 1 36 1 4 29 111 Grand totals 1 80 3 188 1 7 78 761 Expected 1 : 9 8-1 72-9 19-1 168-9 08 72 83-9 755-1 See text Expected, con- antselection 3-7 7-3 4-5 186-5 2-8 5-2 91-4 747-6 -71 * st and that the direction of the selection could change within a few days. These changes did not seem to be caused by alterations in the colour of the background, which - remained apparently constant throughout the period of • study. It remains to enquire whether the differences, between the two experiments are related to the frequencies of the colours. On the basis of the grand totals we can calculate for each species the expected numbers of prey eaten, assuming that the selection remained constant throughout the period of study. If this were so the cross-product - ratio = No. of green offered x No. of brown eaten No. of brown offered x No. of green eaten should be the same in both experiments. The expected numbers based on this assumption are given in the second row below the grand totals. It can be seen that in fact all four species took a larger number of common forms than expected and a smaller number of rare ones. Thus • the birds appeared to select in a frequency-dependent manner. Direct comparison of observed and expected numbers suggests that the effect is highly significant ((X) °/n=20- l9 P<(<0.0001). This comparison, however, is not legitimate because of heterogeneity within the series. In experiment 2, for example, the house sparrows took the - 297 -

Table 2. NUMBERS OF BAITS TAKEN DURING INDIVIDUAL VISITS ON AUGUST 2, 1966 No. of baits taken Visit G B 1 4 0 2 7 0 3 3 0 4 3 0 5 9 0 6 5 0 7 5 0 8 2 0 9 0 10 10 24 0 11 - 3 0 12 5 0 13 1 0 14 0 12 Total 71 22 two colours in different proportions on different days (Xis) = 5915 P.z0001). Even where there is no significant heterogeneity between days, as with the results from the blackbirds, there is strong evidence that the proportions differed from visit to visit. Table 2 shows the numbers of green and brown baits taken by blackbirds on August 2, 1966. Each visit by an individual bird is recorded separately. The data are not amenable to analysis by chi-squared, but it is clear that they are very significantly heterogeneous. If the birds were taking the baits at random, but encountering them in the proportions of seventy-one greens to twenty- two browns, then the probability of a bird taking twelve browns in succession (as on visit 14) would be less than 1 in 10. When this heterogeneity is taken into account, the frequency-dependent effects cannot be regarded as formally significant. They nevertheless fulfil predictions based on the hypothesis of apostatic selection. This hypothesis is further strengthened by a second set of experiments in which "new" birds were familiarized with one colour before they were allowed to feed on a population containing both in equal proportions. The experiments were carried out in a garden near Dalkeith, Midlothian. During 6 days in September 1966, approximately 2,000 green baits were scattered on a background of dark brown soil marked out in a rectangular grid of forty one-metre squares. A population of forty greens and forty browns was then presented to the birds on 3 consecutive days. The greens were somewhat more conspicuous to the human eye. After a lapse of 1 month, the familiarization was repeated with brown baits, after which the birds were again exposed to a- population containing forty of each colour. During the second experiment dunnocks, house sparrows and robins were feeding at the same time as the blackbirds. It was, however, only practicable to observe the blackbirds. Predation by "small birds" was deduced from an examination of the - baits at half hour intervals. The results are given in Table 3. In each case the blackbirds took more of the familiar colour. The trend is highly significant despite the hetero- - .298 -

Table 3. DAILY TOTALS OF BAITS TAKEN AFTER FAMILIARIZATION WITS ONE COLOUR ALONE: SOIL BACKGROUND Hours of Observed predation Date observation Blackbirds "Small birds" (BST) (8+) (many) C B C B Familiarized 6.9.66 14.30-21.00 50 (9) 9 (1) - - with greens 7:9.66 11.15-19.00 111 (13) 14 (0) - - 8.9.66 6.20-18.22 153 (32) 75 (8) - - Totals 314 (53) 98 (9) - - Familiarized with browns 15.10.66 8.25-16.05 52 (8) 103 (20) 110 254 For an explanation of the numbers in parentheses, see text. geneity between visits. This point is well r!lade by the figures in parentheses in Table 3, which record the numbers of visits during which the specified colour was taken in excess. For each day these figures depart significantly in the expected direction from a 1: 1 ratio. After familiarization with greens thre was a significant increase in the percentage of browns taken by blackbirds over every five visits (P <0-01, Spearman's rank correlation test), suggesting that the birds were gradually recognizing the browns as acceptable food. During the single day after familiarization with browns there was no corresponding increase in the number of greens taken by blackbirds, but the "small birds'.' did show such a trend (P < 0.01). The heterogeneity between visits, found in both sets of experiments, suggests that individual birds may differ in their preferences ("specific searching images") for particular colours. This suggestion has been confirmed for greens and browns by observations on colour-ringed blackbirds (Allen, in preparation). The second set of experiments indicates that these preferences, can change systematically as a result of frequent encounters with a particular colour. Such behaviour would be expected to give 'rise to the frequency-dependent effects shown in Table 1. In summary, ground-feeding passerines, which are important predators of many polymorphic species of insects and molluscs, appear to hunt in a manner that would tend actively to maintain colour polymorphisms in their prey. We are grateful to Dr John Godfrey for critically reading the manuscript, and to the Science Research Council for financial support. JOHN A. ALLEN BRYAN CLARKE Department of Zoology, University of Edinburgh. Received July 10, 1968. 'Tinbergen,'L., Arch8. Néeri. Zool., 13, 265 (1960). Clarke, B., In Taxonomy and Geography (edit. by Nichols, D.). 47 (Syste- matics Association, Oxford, 1962). 'Cain, A. J., and Sheppard, P. H., Genetics, 39, 89 (1954). 'Clarke, B., and ODonald, P., Heredity, 19, 201 (1964). 'Clarke, B., Evolution, 18, 364 (1964). 'Croze, H., thesis, Univ. Oxford (1967). Allen, J. A., thesis, Univ. Edinburgh (1967). The Munsdll Book of Color (Mansell Color Co. Inc., Baltimore, 1966). Printed in Great Britain by Fisher. Knight & Co.. Ltd.. St. Albans. - 299 -

(Reprinted, with corrections of publishers errors, from Nature, 237: 348-349 (1972)).

Evidence for Stabilizing and ic Selection by Wild Blackbirds

WHEN predators concentrate on rare varieties of prey species as previous work has suggested 1 there is a tendency towards the elimination of variability. This type of selectbn is stabilizing 5 . When predators choose the most common varieties, the selection is apostotic 6 ,and may result in the maintenance of variability69 . My results (ref. 10 and in preparation) with wild blackbirds (Turdus merula L.) and artificial prey suggest that selection may be stabilizing or apostatic according to the density of the prey. The prey were cylindrical (0.7 cm long x 0.7 cm diameter) lard-and-flour baits coloured green or brown 7 . In a first set of experiments,, during April and May 1968 9 closely packed baits were distributed at random in two circular (19 cm diameter) metal sieves. One sieve contained 30 greens and 270 browns, the other 270 greens and 30 browns. They were placed 1 m apart on the lawn at the Department of Zoology, University of Edinburgh. Predation was confined to a pair of blackbirds and the number of baits taken was deduced from a frequent examination of the sieves. Eaten baits were replaced to ensure that the 9:1 ratios were kept constant.

The populations were presented daily for 2 to. 4 h and their relative positions were frequently altered throughout the 37 days of the study. Table 1 gives the grand totals of baits taken and the results expected on the basis of no selection. The blackbirds clearly - 300 -

Table 1. Grand Totals of Baits taken from 9:1 Populations presented at Maximum Density

Dates of Population 1 Population 2 presentation 9 green: 1 brown 1 green: 9 brown (1968) G B G B April 22-May 27 1,606 592 73 942 (1 9 97802) (2198) (1685) (1,516.5)

Population size = 300 baits. Figures in parentheses represent the expectation if the baits were taken in proportion to their. frequencies. G, green B, brown. - 301 -

tended to select the rarer colours. This result is statistically highly significant.

In a second set of experiments, carried out between 1966 and 1969, populations were presented at densities of two baits per

square metre on lawns and grass fields near Edinburgh. Including

a preliminary study 7 , eleven experiments gave data for blackbirds alone, and in one case predation by a single songthrush (Turdus

philoinelos Hart.) is included. The populations usually consisted

of 200 baits distributed at random in grids of 100 one-metre squares,

each metre square containing two baits. In one experiment -100 baits were randomly distributed in a grid of 50 one-metre squares.

The populations of three experiments were observed continuously.

Others were sometimes left unattended for short periods, after which they were examined to determine the number of baits removed. In each experiment the spatial distribution of the baits was altered

daily and eaten baits' were replaced frequently. In five experiments the populations were first presented with greens nine times as common as browns, followed by populations with browns nine times commoner than greens. In the remaining six experiments the populations were presented in the reverse sequence. The experiments varied in length, but most lasted for about 2 weeks.

The overall results are given in Table 2. We must enquire whether selection was related to the frequencies of the colours. If selection had remained constant throughout a given experiment the cross-product ratio:

No. of green offered x No. of brown eaten No. of brown offered x No. of green eaten - 302 -

Table 2, Grand Totals of Baits taken from 9:1 Populations presented at Densities of 2/rn 2

Nos. of 9 green : 1 brown >1 green : 9 brown Expt. blackbirds

1 6 + •' 3513 86 1 80 238 NS (3138.9) (91.1) (3.7) (77.3).

5.2 13 353 56 U. 278 4,32 <0.01 (343.0) (66.0) (17.13) (271.6)

5.3 2 . 231 130 7 299 8.60 <0.01 (217.0) (513.0) (113.5) (291.5)

5.4 2 192 39 6 348 10082 <0.01 (175.6) (55.4) - (13.6) (3130.4).

6.2 13 . .519 148 .5 315 8.13 <0.01 (601.3) (165.7) (13.6) (3013.4)

Nos. of . I Veen i 9 brown > 9 green : 1 brown Expt. blackbirds B G. X2Q) P

2.1 2 and 1 1 312 92 . 327 0.01 NS songthrush (1.1) (311.9) (91.5) (327.5)

13 2 + 0 232 117 . 74 7.46 <0.05 (3.3) (228.7) (103.1) (81.9)

33 10 . 2 734, 63 102 5.413 <0.02 (4,0) (732.0) . (505) (11.5)

5.1 13 3 368 , . 372 90 15.18 <0.001 (13.5) (357.5) (348.0) (1113.0)

5.5' .5 :• 3 350 380 81 10.25 <0.001 (14.2) (338.8) (356.1.) (1013.9)

6.1 4 13 606 411 131 .20.72 <0,001 (16.5) (593.6) (375.13) (166.6)

Populations size 200 baits except where shown. Figures in parentheses represent the expectation assuming constant selection. 3, green; B, brown; NS, not significant. Continuously observed. + The two 1191" populations were separated by a period of presentation of 111:1" populations. j Population size = 100 baits. - 303 - should be the same during both parts of the experiment The expected grand totals based on this assumption are given in Table

2. In nine cases the deviations from expected were statistically significant, but such direct comparisons of observed and expected numbers are not strictly legitimate, because of heterogeneity within the experiments7' 10. In all eleven experiments the blackbirds took a larger number of common forms than expected and a smaller number of rare ones. This proportion of "SUCC5SSGS" is statistically significant (P c 0.001, binomial test). The results therefore appear to be consistent, despite heterogeneity. To summarize, selection by wild blackbirds seems to be stabilizing when baits are at a very high density and apostatic when the density is lower. Predation would presumably be random at some intermediate density. At very high prey densities predators may tend to form "specific searching images" for uncommon varieties, because they stand out against a background of the common sorts2 ,4 or because they are oddls 3 . When prey are at low densities predators will have to search more actively and may profit by the formation of specific searching images for common varieties 6 . Predation may be random at very low prey densities". What constitutes "very high", "low", or "very low" densities of prey in a given environment will depend on the particular species of predator and prey. I thank Professor Bryan Clarke and Dr. D.T.Parkin for critically reading my original draft. This work was carried out at the Department of Zoology, University of Edinburgh, and was supported by the Science Research Council. JOHN A. ALLEN Department of Zoology, P0 Box 35064, University of Dar es Salaam, Dar es Salaam, Tanzania Received December 6, 1971: revised March 1, 1972. - 304 -

1Mueller, H.C., Nature 0 233, 345 (1971).

2Pielowski, Z. Ekologia Polska (A) 0 IX, No.11 (1961).

3 Salt, G.W. Ecol. Mon. 37 113 (1967). Pough, F.H., thesis, Amherst College (1964).

5Schmalhausen, 1.1., Factors of Evolution (Blaldston, Philadelphia, 1949).

6Clarke, B., in Taxonomy and Geogra (edit, by Nichols, D.), 47 (Systematics Association, Oxford, .1962). ?Allen, J,A., and Clarke, B., Nature, 220, 501 (1968).

8Clarke, B., Heredity, 24, 347 (1969).

9Clarke, B., and O'Donald, P., Heredity, 19, 201 (1964); 10 Allen, J.A., thesis, Univ. Edinburgh (1972).

11 Tinbergen, L., Archs.Nerl.Zoo1., 13, 265 (1960).